llm.rb 14.0.0 → 15.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (92) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +396 -1826
  3. data/README.md +596 -490
  4. data/bin/llm.rb +148 -72
  5. data/data/alibaba.json +1999 -0
  6. data/data/anthropic.json +205 -205
  7. data/data/bedrock.json +2171 -2114
  8. data/data/deepinfra.json +1143 -938
  9. data/data/deepseek.json +4 -5
  10. data/data/google.json +691 -779
  11. data/data/mistral.json +450 -450
  12. data/data/moonshot.json +100 -100
  13. data/data/openai.json +976 -976
  14. data/data/xai.json +193 -116
  15. data/data/zai.json +187 -187
  16. data/{resources → docs}/deepdive/advanced/cancellation.md +2 -2
  17. data/{resources → docs}/deepdive/advanced/compaction.md +5 -3
  18. data/{resources → docs}/deepdive/advanced/context.md +13 -0
  19. data/{resources → docs}/deepdive/advanced/guard.md +1 -1
  20. data/{resources/deepdive/fundamentals → docs/deepdive/features}/concurrency.md +6 -0
  21. data/{resources/deepdive/fundamentals → docs/deepdive/features}/repl.md +56 -2
  22. data/{resources → docs}/deepdive/fundamentals/agents.md +52 -1
  23. data/docs/deepdive/fundamentals/providers.md +159 -0
  24. data/{resources → docs}/deepdive/fundamentals/skills.md +5 -0
  25. data/{resources → docs}/deepdive/fundamentals/stream.md +36 -3
  26. data/{resources → docs}/deepdive/fundamentals/tools.md +87 -23
  27. data/{resources/deepdive/everything_else → docs/deepdive/reference}/cost.md +20 -10
  28. data/docs/deepdive/reference/model_registry.md +271 -0
  29. data/{resources/deepdive/advanced → docs/deepdive/reference}/tracer.md +7 -0
  30. data/{resources → docs}/deepdive.md +35 -27
  31. data/lib/llm/a2a/transport/http.rb +1 -1
  32. data/lib/llm/active_record/acts_as_llm.rb +19 -5
  33. data/lib/llm/agent.rb +62 -5
  34. data/lib/llm/context.rb +93 -46
  35. data/lib/llm/cost.rb +110 -51
  36. data/lib/llm/error.rb +7 -0
  37. data/lib/llm/function/array.rb +1 -1
  38. data/lib/llm/function/fork/task.rb +14 -1
  39. data/lib/llm/function/sequential/group.rb +20 -13
  40. data/lib/llm/function/sequential/task.rb +1 -8
  41. data/lib/llm/function.rb +5 -4
  42. data/lib/llm/message.rb +5 -4
  43. data/lib/llm/provider.rb +7 -0
  44. data/lib/llm/providers/alibaba/error_handler.rb +34 -0
  45. data/lib/llm/providers/alibaba/request_adapter.rb +13 -0
  46. data/lib/llm/providers/alibaba.rb +93 -0
  47. data/lib/llm/providers/anthropic.rb +0 -1
  48. data/lib/llm/providers/bedrock.rb +8 -1
  49. data/lib/llm/providers/deepseek/request_adapter.rb +2 -33
  50. data/lib/llm/providers/google.rb +0 -1
  51. data/lib/llm/providers/ollama.rb +0 -1
  52. data/lib/llm/providers/openai/responses.rb +0 -1
  53. data/lib/llm/providers/openai/schema.rb +37 -0
  54. data/lib/llm/providers/openai.rb +1 -2
  55. data/lib/llm/registry/model.rb +186 -0
  56. data/lib/llm/registry.rb +45 -14
  57. data/lib/llm/repl/bar.rb +11 -13
  58. data/lib/llm/repl/buffer.rb +1 -1
  59. data/lib/llm/repl/color.rb +8 -1
  60. data/lib/llm/repl/command.rb +12 -0
  61. data/lib/llm/repl/commands/model.rb +39 -0
  62. data/lib/llm/repl/input/cache.rb +45 -0
  63. data/lib/llm/repl/input/char.rb +2 -2
  64. data/lib/llm/repl/input.rb +86 -23
  65. data/lib/llm/repl/markdown.rb +25 -1
  66. data/lib/llm/repl/node.rb +7 -0
  67. data/lib/llm/repl/status.rb +17 -3
  68. data/lib/llm/repl/window.rb +91 -11
  69. data/lib/llm/repl.rb +18 -8
  70. data/lib/llm/sequel/plugin.rb +19 -5
  71. data/lib/llm/skill.rb +21 -8
  72. data/lib/llm/stream.rb +27 -0
  73. data/lib/llm/tool.rb +3 -5
  74. data/lib/llm/tools/rg.rb +2 -1
  75. data/lib/llm/transport/curb.rb +23 -3
  76. data/lib/llm/usage.rb +155 -9
  77. data/lib/llm/version.rb +1 -1
  78. data/lib/llm.rb +111 -29
  79. data/llm.gemspec +16 -11
  80. metadata +89 -35
  81. /data/{resources → docs}/deepdive/advanced/transformer.md +0 -0
  82. /data/{resources → docs}/deepdive/advanced/transports.md +0 -0
  83. /data/{resources/deepdive/fundamentals → docs/deepdive/features}/builtin_tools.md +0 -0
  84. /data/{resources/deepdive/fundamentals → docs/deepdive/features}/database.md +0 -0
  85. /data/{resources/deepdive/fundamentals → docs/deepdive/features}/embeddings.md +0 -0
  86. /data/{resources → docs}/deepdive/fundamentals/schema.md +0 -0
  87. /data/{resources/deepdive/everything_else → docs/deepdive/media}/audio.md +0 -0
  88. /data/{resources/deepdive/everything_else → docs/deepdive/media}/images.md +0 -0
  89. /data/{resources/deepdive/everything_else → docs/deepdive/media}/ocr.md +0 -0
  90. /data/{resources → docs}/deepdive/protocols/a2a.md +0 -0
  91. /data/{resources → docs}/deepdive/protocols/mcp.md +0 -0
  92. /data/{resources/deepdive/everything_else → docs/deepdive/reference}/object.md +0 -0
data/data/deepinfra.json CHANGED
@@ -7,19 +7,19 @@
7
7
  "name": "Deep Infra",
8
8
  "doc": "https://deepinfra.com/models",
9
9
  "models": {
10
- "nvidia/Llama-3.3-Nemotron-Super-49B-v1.5": {
11
- "id": "nvidia/Llama-3.3-Nemotron-Super-49B-v1.5",
12
- "name": "Llama 3.3 Nemotron Super 49B v1.5",
13
- "description": "Nemotron model for efficient reasoning, coding, and specialized AI agents",
14
- "family": "nemotron",
10
+ "tencent/Hy3": {
11
+ "id": "tencent/Hy3",
12
+ "name": "Hy3",
13
+ "description": "Tencent Hy reasoning model for coding, instruction following, and agent tasks",
14
+ "family": "Hy",
15
15
  "attachment": false,
16
16
  "reasoning": true,
17
17
  "reasoning_options": [],
18
18
  "tool_call": true,
19
19
  "structured_output": true,
20
20
  "temperature": true,
21
- "release_date": "2025-07-25",
22
- "last_updated": "2025-07-25",
21
+ "release_date": "2026-07-06",
22
+ "last_updated": "2026-07-06",
23
23
  "modalities": {
24
24
  "input": [
25
25
  "text"
@@ -30,34 +30,42 @@
30
30
  },
31
31
  "open_weights": true,
32
32
  "limit": {
33
- "context": 131072,
34
- "output": 131072
33
+ "context": 262144,
34
+ "output": 64000
35
35
  },
36
- "status": "deprecated",
37
36
  "cost": {
38
- "input": 0.4,
39
- "output": 0.4
37
+ "input": 0.14,
38
+ "output": 0.58,
39
+ "cache_read": 0.035
40
40
  }
41
41
  },
42
- "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning": {
43
- "id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
44
- "name": "Nemotron 3 Nano Omni 30B A3B Reasoning",
45
- "description": "Open Nemotron omni model combining reasoning with text, vision, and audio",
46
- "family": "nemotron",
42
+ "XiaomiMiMo/MiMo-V2.5": {
43
+ "id": "XiaomiMiMo/MiMo-V2.5",
44
+ "name": "MiMo-V2.5",
45
+ "description": "Open MiMo model for multimodal coding agents and long-context automation",
46
+ "family": "mimo",
47
47
  "attachment": true,
48
48
  "reasoning": true,
49
- "reasoning_options": [],
49
+ "reasoning_options": [
50
+ {
51
+ "type": "toggle"
52
+ }
53
+ ],
50
54
  "tool_call": true,
55
+ "interleaved": {
56
+ "field": "reasoning_content"
57
+ },
51
58
  "structured_output": true,
52
59
  "temperature": true,
53
- "release_date": "2026-04-28",
54
- "last_updated": "2026-04-28",
60
+ "knowledge": "2024-12",
61
+ "release_date": "2026-04-22",
62
+ "last_updated": "2026-04-22",
55
63
  "modalities": {
56
64
  "input": [
57
65
  "text",
58
66
  "image",
59
- "video",
60
- "audio"
67
+ "audio",
68
+ "video"
61
69
  ],
62
70
  "output": [
63
71
  "text"
@@ -66,20 +74,20 @@
66
74
  "open_weights": true,
67
75
  "limit": {
68
76
  "context": 262144,
69
- "output": 65536
77
+ "output": 16384
70
78
  },
71
- "status": "deprecated",
72
79
  "cost": {
73
- "input": 0.2,
74
- "output": 0.8
80
+ "input": 0.4,
81
+ "output": 2,
82
+ "cache_read": 0.08
75
83
  }
76
84
  },
77
- "nvidia/Nemotron-3-Nano-30B-A3B": {
78
- "id": "nvidia/Nemotron-3-Nano-30B-A3B",
79
- "name": "Nemotron 3 Nano 30B A3B",
80
- "description": "Small Nemotron 3 MoE for efficient coding, math, and long-context agents",
81
- "family": "nemotron",
82
- "attachment": false,
85
+ "XiaomiMiMo/MiMo-V2.5-Pro": {
86
+ "id": "XiaomiMiMo/MiMo-V2.5-Pro",
87
+ "name": "MiMo-V2.5-Pro",
88
+ "description": "Stronger MiMo Pro tier for multimodal reasoning and coding-agent execution",
89
+ "family": "mimo",
90
+ "attachment": true,
83
91
  "reasoning": true,
84
92
  "reasoning_options": [
85
93
  {
@@ -87,12 +95,18 @@
87
95
  }
88
96
  ],
89
97
  "tool_call": true,
98
+ "interleaved": {
99
+ "field": "reasoning_content"
100
+ },
101
+ "structured_output": true,
90
102
  "temperature": true,
91
- "release_date": "2025-12-15",
92
- "last_updated": "2025-12-15",
103
+ "knowledge": "2024-12",
104
+ "release_date": "2026-04-22",
105
+ "last_updated": "2026-04-22",
93
106
  "modalities": {
94
107
  "input": [
95
- "text"
108
+ "text",
109
+ "audio"
96
110
  ],
97
111
  "output": [
98
112
  "text"
@@ -100,36 +114,30 @@
100
114
  },
101
115
  "open_weights": true,
102
116
  "limit": {
103
- "context": 262144,
104
- "output": 262144
117
+ "context": 1048576,
118
+ "output": 16384
105
119
  },
106
120
  "cost": {
107
- "input": 0.05,
108
- "output": 0.2,
109
- "cache_read": 0.025
121
+ "input": 1,
122
+ "output": 3,
123
+ "cache_read": 0.2
110
124
  }
111
125
  },
112
- "google/gemma-4-26B-A4B-it": {
113
- "id": "google/gemma-4-26B-A4B-it",
114
- "name": "Gemma 4 26B A4B IT",
115
- "description": "Open Gemma instruction model for efficient chat and self-hosted deployments",
116
- "family": "gemma",
117
- "attachment": true,
126
+ "MiniMaxAI/MiniMax-M2.7": {
127
+ "id": "MiniMaxAI/MiniMax-M2.7",
128
+ "name": "MiniMax-M2.7",
129
+ "description": "Open MiniMax flagship for coding agents, office automation, and complex environments",
130
+ "family": "minimax",
131
+ "attachment": false,
118
132
  "reasoning": true,
119
- "reasoning_options": [
120
- {
121
- "type": "toggle"
122
- }
123
- ],
133
+ "reasoning_options": [],
124
134
  "tool_call": true,
125
- "structured_output": true,
126
135
  "temperature": true,
127
- "release_date": "2026-04-02",
128
- "last_updated": "2026-04-02",
136
+ "release_date": "2026-03-18",
137
+ "last_updated": "2026-03-18",
129
138
  "modalities": {
130
139
  "input": [
131
- "text",
132
- "image"
140
+ "text"
133
141
  ],
134
142
  "output": [
135
143
  "text"
@@ -137,36 +145,33 @@
137
145
  },
138
146
  "open_weights": true,
139
147
  "limit": {
140
- "context": 262144,
141
- "output": 32768
148
+ "context": 196608,
149
+ "output": 131072
142
150
  },
143
151
  "cost": {
144
- "input": 0.07,
145
- "output": 0.34
152
+ "input": 0.25,
153
+ "output": 1,
154
+ "cache_read": 0.05
146
155
  }
147
156
  },
148
- "google/gemma-4-E4B-it": {
149
- "id": "google/gemma-4-E4B-it",
150
- "name": "Gemma 4 E4B IT",
151
- "description": "Open Gemma instruction model for efficient chat and self-hosted deployments",
152
- "family": "gemma",
157
+ "MiniMaxAI/MiniMax-M3": {
158
+ "id": "MiniMaxAI/MiniMax-M3",
159
+ "name": "MiniMax-M3",
160
+ "description": "MiniMax multimodal model for long-context coding, perception, and agent planning",
161
+ "family": "minimax",
153
162
  "attachment": true,
154
163
  "reasoning": true,
155
- "reasoning_options": [
156
- {
157
- "type": "toggle"
158
- }
159
- ],
164
+ "reasoning_options": [],
160
165
  "tool_call": true,
161
166
  "structured_output": true,
162
167
  "temperature": true,
163
- "release_date": "2026-04-02",
164
- "last_updated": "2026-04-02",
168
+ "release_date": "2026-06-01",
169
+ "last_updated": "2026-06-01",
165
170
  "modalities": {
166
171
  "input": [
167
172
  "text",
168
173
  "image",
169
- "audio"
174
+ "video"
170
175
  ],
171
176
  "output": [
172
177
  "text"
@@ -174,36 +179,34 @@
174
179
  },
175
180
  "open_weights": true,
176
181
  "limit": {
177
- "context": 131072,
178
- "output": 8192
182
+ "context": 524288,
183
+ "output": 128000
179
184
  },
180
185
  "cost": {
181
- "input": 0.02,
182
- "output": 0.1
186
+ "input": 0.28,
187
+ "output": 1.1,
188
+ "cache_read": 0.056
183
189
  }
184
190
  },
185
- "google/gemma-4-31B-it": {
186
- "id": "google/gemma-4-31B-it",
187
- "name": "Gemma 4 31B IT",
188
- "description": "Largest Gemma 4 instruction model for open, self-hosted chat and reasoning",
189
- "family": "gemma",
190
- "attachment": true,
191
+ "MiniMaxAI/MiniMax-M2.5": {
192
+ "id": "MiniMaxAI/MiniMax-M2.5",
193
+ "name": "MiniMax M2.5",
194
+ "description": "MiniMax model for chat, coding, office work, and agentic tasks",
195
+ "family": "minimax",
196
+ "attachment": false,
191
197
  "reasoning": true,
192
- "reasoning_options": [
193
- {
194
- "type": "toggle"
195
- }
196
- ],
198
+ "reasoning_options": [],
197
199
  "tool_call": true,
198
- "structured_output": true,
200
+ "interleaved": {
201
+ "field": "reasoning_content"
202
+ },
199
203
  "temperature": true,
200
- "release_date": "2026-04-02",
201
- "last_updated": "2026-04-02",
204
+ "knowledge": "2025-06",
205
+ "release_date": "2026-02-12",
206
+ "last_updated": "2026-02-12",
202
207
  "modalities": {
203
208
  "input": [
204
- "text",
205
- "image",
206
- "video"
209
+ "text"
207
210
  ],
208
211
  "output": [
209
212
  "text"
@@ -211,31 +214,32 @@
211
214
  },
212
215
  "open_weights": true,
213
216
  "limit": {
214
- "context": 262144,
215
- "output": 32768
217
+ "context": 196608,
218
+ "output": 131072
216
219
  },
220
+ "status": "deprecated",
217
221
  "cost": {
218
- "input": 0.13,
219
- "output": 0.38
222
+ "input": 0.15,
223
+ "output": 1.15,
224
+ "cache_read": 0.03
220
225
  }
221
226
  },
222
- "thinkingmachines/Inkling-Small": {
223
- "id": "thinkingmachines/Inkling-Small",
224
- "name": "Inkling Small",
225
- "description": "Multimodal MoE reasoning model (276B total, 12B active) for text, image, and audio",
226
- "family": "ling",
227
- "attachment": true,
227
+ "nvidia/Llama-3.3-Nemotron-Super-49B-v1.5": {
228
+ "id": "nvidia/Llama-3.3-Nemotron-Super-49B-v1.5",
229
+ "name": "Llama 3.3 Nemotron Super 49B v1.5",
230
+ "description": "Nemotron model for efficient reasoning, coding, and specialized AI agents",
231
+ "family": "nemotron",
232
+ "attachment": false,
228
233
  "reasoning": true,
229
234
  "reasoning_options": [],
230
235
  "tool_call": true,
236
+ "structured_output": true,
231
237
  "temperature": true,
232
- "release_date": "2026-07-30",
233
- "last_updated": "2026-07-30",
238
+ "release_date": "2025-07-25",
239
+ "last_updated": "2025-07-25",
234
240
  "modalities": {
235
241
  "input": [
236
- "text",
237
- "image",
238
- "audio"
242
+ "text"
239
243
  ],
240
244
  "output": [
241
245
  "text"
@@ -243,32 +247,34 @@
243
247
  },
244
248
  "open_weights": true,
245
249
  "limit": {
246
- "context": 524288,
247
- "output": 1048576
250
+ "context": 131072,
251
+ "output": 131072
248
252
  },
253
+ "status": "deprecated",
249
254
  "cost": {
250
- "input": 0.45,
251
- "output": 1.2,
252
- "cache_read": 0.1
255
+ "input": 0.4,
256
+ "output": 0.4
253
257
  }
254
258
  },
255
- "thinkingmachines/Inkling": {
256
- "id": "thinkingmachines/Inkling",
257
- "name": "Inkling",
258
- "description": "Multimodal MoE reasoning model (975B total, 41B active) for text, image, and audio",
259
- "family": "ling",
260
- "attachment": true,
259
+ "nvidia/Nemotron-3-Nano-30B-A3B": {
260
+ "id": "nvidia/Nemotron-3-Nano-30B-A3B",
261
+ "name": "Nemotron 3 Nano 30B A3B",
262
+ "description": "Small Nemotron 3 MoE for efficient coding, math, and long-context agents",
263
+ "family": "nemotron",
264
+ "attachment": false,
261
265
  "reasoning": true,
262
- "reasoning_options": [],
266
+ "reasoning_options": [
267
+ {
268
+ "type": "toggle"
269
+ }
270
+ ],
263
271
  "tool_call": true,
264
272
  "temperature": true,
265
- "release_date": "2026-07-15",
266
- "last_updated": "2026-07-15",
273
+ "release_date": "2025-12-15",
274
+ "last_updated": "2025-12-15",
267
275
  "modalities": {
268
276
  "input": [
269
- "text",
270
- "image",
271
- "audio"
277
+ "text"
272
278
  ],
273
279
  "output": [
274
280
  "text"
@@ -276,39 +282,34 @@
276
282
  },
277
283
  "open_weights": true,
278
284
  "limit": {
279
- "context": 524288,
280
- "output": 1048576
285
+ "context": 262144,
286
+ "output": 262144
281
287
  },
282
288
  "cost": {
283
- "input": 0.95,
284
- "output": 4.05,
285
- "cache_read": 0.16
289
+ "input": 0.05,
290
+ "output": 0.2,
291
+ "cache_read": 0.025
286
292
  }
287
293
  },
288
- "zai-org/GLM-5": {
289
- "id": "zai-org/GLM-5",
290
- "name": "GLM-5",
291
- "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
292
- "family": "glm",
293
- "attachment": false,
294
+ "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning": {
295
+ "id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
296
+ "name": "Nemotron 3 Nano Omni 30B A3B Reasoning",
297
+ "description": "Open Nemotron omni model combining reasoning with text, vision, and audio",
298
+ "family": "nemotron",
299
+ "attachment": true,
294
300
  "reasoning": true,
295
- "reasoning_options": [
296
- {
297
- "type": "toggle"
298
- }
299
- ],
301
+ "reasoning_options": [],
300
302
  "tool_call": true,
301
- "interleaved": {
302
- "field": "reasoning_content"
303
- },
304
303
  "structured_output": true,
305
304
  "temperature": true,
306
- "knowledge": "2025-12",
307
- "release_date": "2026-02-12",
308
- "last_updated": "2026-02-12",
305
+ "release_date": "2026-04-28",
306
+ "last_updated": "2026-04-28",
309
307
  "modalities": {
310
308
  "input": [
311
- "text"
309
+ "text",
310
+ "image",
311
+ "video",
312
+ "audio"
312
313
  ],
313
314
  "output": [
314
315
  "text"
@@ -316,35 +317,36 @@
316
317
  },
317
318
  "open_weights": true,
318
319
  "limit": {
319
- "context": 202752,
320
- "output": 16384
320
+ "context": 262144,
321
+ "output": 65536
321
322
  },
323
+ "status": "deprecated",
322
324
  "cost": {
323
- "input": 0.6,
324
- "output": 2.08,
325
- "cache_read": 0.12
325
+ "input": 0.2,
326
+ "output": 0.8
326
327
  }
327
328
  },
328
- "zai-org/GLM-4.7-Flash": {
329
- "id": "zai-org/GLM-4.7-Flash",
330
- "name": "GLM-4.7-Flash",
331
- "description": "Efficient GLM model for fast reasoning, coding, and agent workflows",
332
- "family": "glm-flash",
333
- "attachment": false,
329
+ "google/gemma-4-26B-A4B-it": {
330
+ "id": "google/gemma-4-26B-A4B-it",
331
+ "name": "Gemma 4 26B A4B IT",
332
+ "description": "Open Gemma instruction model for efficient chat and self-hosted deployments",
333
+ "family": "gemma",
334
+ "attachment": true,
334
335
  "reasoning": true,
335
- "reasoning_options": [],
336
+ "reasoning_options": [
337
+ {
338
+ "type": "toggle"
339
+ }
340
+ ],
336
341
  "tool_call": true,
337
- "interleaved": {
338
- "field": "reasoning_content"
339
- },
340
342
  "structured_output": true,
341
343
  "temperature": true,
342
- "knowledge": "2025-04",
343
- "release_date": "2026-01-19",
344
- "last_updated": "2026-01-19",
344
+ "release_date": "2026-04-02",
345
+ "last_updated": "2026-04-02",
345
346
  "modalities": {
346
347
  "input": [
347
- "text"
348
+ "text",
349
+ "image"
348
350
  ],
349
351
  "output": [
350
352
  "text"
@@ -352,47 +354,36 @@
352
354
  },
353
355
  "open_weights": true,
354
356
  "limit": {
355
- "context": 202752,
356
- "output": 16384
357
+ "context": 262144,
358
+ "output": 32768
357
359
  },
358
360
  "cost": {
359
- "input": 0.06,
360
- "output": 0.4,
361
- "cache_read": 0.01
361
+ "input": 0.07,
362
+ "output": 0.34
362
363
  }
363
364
  },
364
- "zai-org/GLM-5.2": {
365
- "id": "zai-org/GLM-5.2",
366
- "name": "GLM-5.2",
367
- "description": "Open flagship GLM for long-horizon coding agents and million-token context work",
368
- "family": "glm",
369
- "attachment": false,
365
+ "google/gemma-4-E4B-it": {
366
+ "id": "google/gemma-4-E4B-it",
367
+ "name": "Gemma 4 E4B IT",
368
+ "description": "Open Gemma instruction model for efficient chat and self-hosted deployments",
369
+ "family": "gemma",
370
+ "attachment": true,
370
371
  "reasoning": true,
371
372
  "reasoning_options": [
372
373
  {
373
374
  "type": "toggle"
374
- },
375
- {
376
- "type": "effort",
377
- "values": [
378
- "low",
379
- "medium",
380
- "high",
381
- "xhigh"
382
- ]
383
375
  }
384
376
  ],
385
377
  "tool_call": true,
386
- "interleaved": {
387
- "field": "reasoning_content"
388
- },
389
378
  "structured_output": true,
390
379
  "temperature": true,
391
- "release_date": "2026-06-13",
392
- "last_updated": "2026-06-13",
380
+ "release_date": "2026-04-02",
381
+ "last_updated": "2026-04-02",
393
382
  "modalities": {
394
383
  "input": [
395
- "text"
384
+ "text",
385
+ "image",
386
+ "audio"
396
387
  ],
397
388
  "output": [
398
389
  "text"
@@ -400,21 +391,20 @@
400
391
  },
401
392
  "open_weights": true,
402
393
  "limit": {
403
- "context": 1048576,
404
- "output": 32768
394
+ "context": 131072,
395
+ "output": 8192
405
396
  },
406
397
  "cost": {
407
- "input": 0.75,
408
- "output": 2.4,
409
- "cache_read": 0.14
398
+ "input": 0.02,
399
+ "output": 0.1
410
400
  }
411
401
  },
412
- "zai-org/GLM-5.1": {
413
- "id": "zai-org/GLM-5.1",
414
- "name": "GLM-5.1",
415
- "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
416
- "family": "glm",
417
- "attachment": false,
402
+ "google/gemma-4-31B-it": {
403
+ "id": "google/gemma-4-31B-it",
404
+ "name": "Gemma 4 31B IT",
405
+ "description": "Largest Gemma 4 instruction model for open, self-hosted chat and reasoning",
406
+ "family": "gemma",
407
+ "attachment": true,
418
408
  "reasoning": true,
419
409
  "reasoning_options": [
420
410
  {
@@ -422,17 +412,15 @@
422
412
  }
423
413
  ],
424
414
  "tool_call": true,
425
- "interleaved": {
426
- "field": "reasoning_content"
427
- },
428
415
  "structured_output": true,
429
416
  "temperature": true,
430
- "knowledge": "2025-04",
431
- "release_date": "2026-04-07",
432
- "last_updated": "2026-04-07",
417
+ "release_date": "2026-04-02",
418
+ "last_updated": "2026-04-02",
433
419
  "modalities": {
434
420
  "input": [
435
- "text"
421
+ "text",
422
+ "image",
423
+ "video"
436
424
  ],
437
425
  "output": [
438
426
  "text"
@@ -440,146 +428,182 @@
440
428
  },
441
429
  "open_weights": true,
442
430
  "limit": {
443
- "context": 202752,
444
- "output": 16384
431
+ "context": 262144,
432
+ "output": 32768
445
433
  },
446
434
  "cost": {
447
- "input": 1.05,
448
- "output": 3.5,
449
- "cache_read": 0.205
435
+ "input": 0.13,
436
+ "output": 0.38
450
437
  }
451
438
  },
452
- "zai-org/GLM-4.6": {
453
- "id": "zai-org/GLM-4.6",
454
- "name": "GLM-4.6",
455
- "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
456
- "family": "glm",
457
- "attachment": false,
439
+ "ByteDance/Seed-2.0-code": {
440
+ "id": "ByteDance/Seed-2.0-code",
441
+ "name": "Seed 2.0 Code",
442
+ "description": "ByteDance Seed coding model for multimodal software engineering and long-running agents",
443
+ "family": "seed",
444
+ "attachment": true,
458
445
  "reasoning": true,
459
446
  "reasoning_options": [
460
447
  {
461
- "type": "toggle"
448
+ "type": "effort",
449
+ "values": [
450
+ "none",
451
+ "low",
452
+ "medium",
453
+ "high"
454
+ ]
462
455
  }
463
456
  ],
464
457
  "tool_call": true,
465
- "interleaved": {
466
- "field": "reasoning_content"
467
- },
468
458
  "structured_output": true,
469
459
  "temperature": true,
470
- "knowledge": "2025-04",
471
- "release_date": "2025-09-30",
472
- "last_updated": "2025-09-30",
460
+ "release_date": "2026-02-14",
461
+ "last_updated": "2026-02-14",
473
462
  "modalities": {
474
463
  "input": [
475
- "text"
464
+ "text",
465
+ "image"
476
466
  ],
477
467
  "output": [
478
468
  "text"
479
469
  ]
480
470
  },
481
- "open_weights": true,
471
+ "open_weights": false,
482
472
  "limit": {
483
- "context": 202752,
473
+ "context": 256000,
484
474
  "output": 131072
485
475
  },
486
476
  "cost": {
487
477
  "input": 0.5,
488
- "output": 2,
489
- "cache_read": 0.1
478
+ "output": 3,
479
+ "cache_read": 0.1,
480
+ "tiers": [
481
+ {
482
+ "input": 1,
483
+ "output": 6,
484
+ "cache_read": 0.2,
485
+ "tier": {
486
+ "type": "context",
487
+ "size": 128000
488
+ }
489
+ }
490
+ ]
490
491
  }
491
492
  },
492
- "zai-org/GLM-4.7": {
493
- "id": "zai-org/GLM-4.7",
494
- "name": "GLM-4.7",
495
- "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
496
- "family": "glm",
497
- "attachment": false,
493
+ "ByteDance/Seed-2.0-pro": {
494
+ "id": "ByteDance/Seed-2.0-pro",
495
+ "name": "Seed 2.0 Pro",
496
+ "description": "Flagship ByteDance Seed 2.0 model for complex multimodal reasoning and long-horizon agent workflows",
497
+ "family": "seed",
498
+ "attachment": true,
498
499
  "reasoning": true,
499
- "reasoning_options": [
500
- {
501
- "type": "toggle"
502
- }
503
- ],
500
+ "reasoning_options": [],
504
501
  "tool_call": true,
505
- "interleaved": {
506
- "field": "reasoning_content"
507
- },
508
502
  "structured_output": true,
509
503
  "temperature": true,
510
- "knowledge": "2025-04",
511
- "release_date": "2025-12-22",
512
- "last_updated": "2025-12-22",
504
+ "release_date": "2026-02-14",
505
+ "last_updated": "2026-02-14",
513
506
  "modalities": {
514
507
  "input": [
515
- "text"
508
+ "text",
509
+ "image"
516
510
  ],
517
511
  "output": [
518
512
  "text"
519
513
  ]
520
514
  },
521
- "open_weights": true,
515
+ "open_weights": false,
522
516
  "limit": {
523
- "context": 202752,
524
- "output": 16384
517
+ "context": 256000,
518
+ "output": 128000
525
519
  },
526
520
  "cost": {
527
- "input": 0.4,
528
- "output": 1.75,
529
- "cache_read": 0.08
521
+ "input": 0.5,
522
+ "output": 3,
523
+ "cache_read": 0.1,
524
+ "tiers": [
525
+ {
526
+ "input": 1,
527
+ "output": 6,
528
+ "cache_read": 0.2,
529
+ "tier": {
530
+ "type": "context",
531
+ "size": 128000
532
+ }
533
+ }
534
+ ]
530
535
  }
531
536
  },
532
- "tencent/Hy3": {
533
- "id": "tencent/Hy3",
534
- "name": "Hy3",
535
- "description": "Tencent Hy reasoning model for coding, instruction following, and agent tasks",
536
- "family": "Hy",
537
- "attachment": false,
537
+ "ByteDance/Seed-2.0-mini": {
538
+ "id": "ByteDance/Seed-2.0-mini",
539
+ "name": "Seed 2.0 Mini",
540
+ "description": "Lightweight ByteDance Seed 2.0 model for low-latency multimodal reasoning and high-volume tasks",
541
+ "family": "seed",
542
+ "attachment": true,
538
543
  "reasoning": true,
539
544
  "reasoning_options": [],
540
545
  "tool_call": true,
541
546
  "structured_output": true,
542
547
  "temperature": true,
543
- "release_date": "2026-07-06",
544
- "last_updated": "2026-07-06",
548
+ "release_date": "2026-02-14",
549
+ "last_updated": "2026-02-14",
545
550
  "modalities": {
546
551
  "input": [
547
- "text"
552
+ "text",
553
+ "image"
548
554
  ],
549
555
  "output": [
550
556
  "text"
551
557
  ]
552
558
  },
553
- "open_weights": true,
559
+ "open_weights": false,
554
560
  "limit": {
555
- "context": 262144,
556
- "output": 64000
561
+ "context": 256000,
562
+ "output": 32000
557
563
  },
558
564
  "cost": {
559
- "input": 0.14,
560
- "output": 0.58,
561
- "cache_read": 0.035
565
+ "input": 0.1,
566
+ "output": 0.4,
567
+ "cache_read": 0.02,
568
+ "tiers": [
569
+ {
570
+ "input": 0.2,
571
+ "output": 0.8,
572
+ "cache_read": 0.2,
573
+ "tier": {
574
+ "type": "context",
575
+ "size": 128000
576
+ }
577
+ }
578
+ ]
562
579
  }
563
580
  },
564
- "Qwen/Qwen3.5-27B": {
565
- "id": "Qwen/Qwen3.5-27B",
566
- "name": "Qwen3.5 27B",
567
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
568
- "family": "qwen",
581
+ "moonshotai/Kimi-K2.5": {
582
+ "id": "moonshotai/Kimi-K2.5",
583
+ "name": "Kimi K2.5",
584
+ "description": "Kimi multimodal agent model for visual understanding, coding, and planning",
585
+ "family": "kimi-k2",
569
586
  "attachment": true,
570
587
  "reasoning": true,
571
- "reasoning_options": [],
588
+ "reasoning_options": [
589
+ {
590
+ "type": "toggle"
591
+ }
592
+ ],
572
593
  "tool_call": true,
594
+ "interleaved": {
595
+ "field": "reasoning_content"
596
+ },
573
597
  "structured_output": true,
574
598
  "temperature": true,
575
- "release_date": "2026-02-23",
576
- "last_updated": "2026-02-23",
599
+ "knowledge": "2025-01",
600
+ "release_date": "2026-01-27",
601
+ "last_updated": "2026-01-27",
577
602
  "modalities": {
578
603
  "input": [
579
604
  "text",
580
605
  "image",
581
- "video",
582
- "audio"
606
+ "video"
583
607
  ],
584
608
  "output": [
585
609
  "text"
@@ -588,26 +612,119 @@
588
612
  "open_weights": true,
589
613
  "limit": {
590
614
  "context": 262144,
591
- "output": 65536
615
+ "output": 32768
592
616
  },
593
617
  "cost": {
594
- "input": 0.26,
595
- "output": 2.6
618
+ "input": 0.45,
619
+ "output": 2.25,
620
+ "cache_read": 0.07
596
621
  }
597
622
  },
598
- "Qwen/Qwen3.5-9B": {
599
- "id": "Qwen/Qwen3.5-9B",
600
- "name": "Qwen3.5 9B",
601
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
602
- "family": "qwen",
623
+ "moonshotai/Kimi-K2.7-Code": {
624
+ "id": "moonshotai/Kimi-K2.7-Code",
625
+ "name": "Kimi K2.7 Code",
626
+ "description": "Coding-focused Kimi model, stronger on long-horizon repo work with less overthinking",
627
+ "family": "kimi-k2",
603
628
  "attachment": true,
604
629
  "reasoning": true,
605
- "reasoning_options": [],
630
+ "reasoning_options": [
631
+ {
632
+ "type": "toggle"
633
+ }
634
+ ],
635
+ "tool_call": true,
636
+ "interleaved": {
637
+ "field": "reasoning_content"
638
+ },
639
+ "structured_output": true,
640
+ "temperature": false,
641
+ "knowledge": "2025-01",
642
+ "release_date": "2026-06-12",
643
+ "last_updated": "2026-06-12",
644
+ "modalities": {
645
+ "input": [
646
+ "text",
647
+ "image",
648
+ "video"
649
+ ],
650
+ "output": [
651
+ "text"
652
+ ]
653
+ },
654
+ "open_weights": true,
655
+ "limit": {
656
+ "context": 262144,
657
+ "output": 262144
658
+ },
659
+ "cost": {
660
+ "input": 0.68,
661
+ "output": 3.4,
662
+ "cache_read": 0.136
663
+ }
664
+ },
665
+ "moonshotai/Kimi-K3": {
666
+ "id": "moonshotai/Kimi-K3",
667
+ "name": "Kimi K3",
668
+ "description": "Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work",
669
+ "family": "kimi-k3",
670
+ "attachment": true,
671
+ "reasoning": true,
672
+ "reasoning_options": [
673
+ {
674
+ "type": "effort",
675
+ "values": [
676
+ "low",
677
+ "high",
678
+ "max"
679
+ ]
680
+ }
681
+ ],
606
682
  "tool_call": true,
607
683
  "structured_output": true,
684
+ "temperature": false,
685
+ "release_date": "2026-07-16",
686
+ "last_updated": "2026-07-16",
687
+ "modalities": {
688
+ "input": [
689
+ "text",
690
+ "image"
691
+ ],
692
+ "output": [
693
+ "text"
694
+ ]
695
+ },
696
+ "open_weights": true,
697
+ "limit": {
698
+ "context": 1048576,
699
+ "output": 131072
700
+ },
701
+ "cost": {
702
+ "input": 2.85,
703
+ "output": 14.25,
704
+ "cache_read": 0.285
705
+ }
706
+ },
707
+ "moonshotai/Kimi-K2.6": {
708
+ "id": "moonshotai/Kimi-K2.6",
709
+ "name": "Kimi K2.6",
710
+ "description": "Kimi multimodal agent model for visual understanding, coding, and planning",
711
+ "family": "kimi-k2",
712
+ "attachment": true,
713
+ "reasoning": true,
714
+ "reasoning_options": [
715
+ {
716
+ "type": "toggle"
717
+ }
718
+ ],
719
+ "tool_call": true,
720
+ "interleaved": {
721
+ "field": "reasoning_content"
722
+ },
723
+ "structured_output": true,
608
724
  "temperature": true,
609
- "release_date": "2026-02-23",
610
- "last_updated": "2026-02-23",
725
+ "knowledge": "2024-04",
726
+ "release_date": "2026-04-21",
727
+ "last_updated": "2026-04-21",
611
728
  "modalities": {
612
729
  "input": [
613
730
  "text",
@@ -620,26 +737,382 @@
620
737
  },
621
738
  "open_weights": true,
622
739
  "limit": {
623
- "context": 262144,
624
- "output": 65536
740
+ "context": 262144,
741
+ "output": 16384
742
+ },
743
+ "cost": {
744
+ "input": 0.75,
745
+ "output": 3.5,
746
+ "cache_read": 0.15
747
+ }
748
+ },
749
+ "openai/gpt-oss-120b": {
750
+ "id": "openai/gpt-oss-120b",
751
+ "name": "GPT OSS 120B",
752
+ "description": "Open-weight GPT model for self-hosted reasoning and instruction-following workloads",
753
+ "family": "gpt-oss",
754
+ "attachment": false,
755
+ "reasoning": true,
756
+ "reasoning_options": [
757
+ {
758
+ "type": "effort",
759
+ "values": [
760
+ "low",
761
+ "medium",
762
+ "high"
763
+ ]
764
+ }
765
+ ],
766
+ "tool_call": true,
767
+ "structured_output": true,
768
+ "temperature": true,
769
+ "release_date": "2025-08-05",
770
+ "last_updated": "2025-08-05",
771
+ "modalities": {
772
+ "input": [
773
+ "text"
774
+ ],
775
+ "output": [
776
+ "text"
777
+ ]
778
+ },
779
+ "open_weights": true,
780
+ "limit": {
781
+ "context": 131072,
782
+ "output": 16384
783
+ },
784
+ "cost": {
785
+ "input": 0.037,
786
+ "output": 0.17
787
+ }
788
+ },
789
+ "openai/gpt-oss-20b": {
790
+ "id": "openai/gpt-oss-20b",
791
+ "name": "GPT OSS 20B",
792
+ "description": "Open-weight GPT model for self-hosted reasoning and instruction-following workloads",
793
+ "family": "gpt-oss",
794
+ "attachment": false,
795
+ "reasoning": true,
796
+ "reasoning_options": [
797
+ {
798
+ "type": "effort",
799
+ "values": [
800
+ "low",
801
+ "medium",
802
+ "high"
803
+ ]
804
+ }
805
+ ],
806
+ "tool_call": true,
807
+ "structured_output": true,
808
+ "temperature": true,
809
+ "release_date": "2025-08-05",
810
+ "last_updated": "2025-08-05",
811
+ "modalities": {
812
+ "input": [
813
+ "text"
814
+ ],
815
+ "output": [
816
+ "text"
817
+ ]
818
+ },
819
+ "open_weights": true,
820
+ "limit": {
821
+ "context": 131072,
822
+ "output": 16384
823
+ },
824
+ "cost": {
825
+ "input": 0.03,
826
+ "output": 0.14
827
+ }
828
+ },
829
+ "deepseek-ai/DeepSeek-V4-Pro": {
830
+ "id": "deepseek-ai/DeepSeek-V4-Pro",
831
+ "name": "DeepSeek V4 Pro",
832
+ "description": "Open MoE flagship with million-token context for coding and long agent runs",
833
+ "family": "deepseek-thinking",
834
+ "attachment": false,
835
+ "reasoning": true,
836
+ "reasoning_options": [
837
+ {
838
+ "type": "toggle"
839
+ },
840
+ {
841
+ "type": "effort",
842
+ "values": [
843
+ "low",
844
+ "medium",
845
+ "high",
846
+ "xhigh"
847
+ ]
848
+ }
849
+ ],
850
+ "tool_call": true,
851
+ "interleaved": {
852
+ "field": "reasoning_content"
853
+ },
854
+ "structured_output": true,
855
+ "temperature": true,
856
+ "knowledge": "2025-05",
857
+ "release_date": "2026-04-24",
858
+ "last_updated": "2026-04-24",
859
+ "modalities": {
860
+ "input": [
861
+ "text"
862
+ ],
863
+ "output": [
864
+ "text"
865
+ ]
866
+ },
867
+ "open_weights": true,
868
+ "limit": {
869
+ "context": 1048576,
870
+ "output": 16384
871
+ },
872
+ "cost": {
873
+ "input": 1.3,
874
+ "output": 2.6,
875
+ "cache_read": 0.1
876
+ }
877
+ },
878
+ "deepseek-ai/DeepSeek-V4-Flash-0731": {
879
+ "id": "deepseek-ai/DeepSeek-V4-Flash-0731",
880
+ "name": "DeepSeek V4 Flash 0731",
881
+ "description": "Official DeepSeek V4 Flash release with enhanced agentic capabilities and integrated DSpark speculative decoding",
882
+ "family": "deepseek-flash",
883
+ "attachment": false,
884
+ "reasoning": true,
885
+ "reasoning_options": [
886
+ {
887
+ "type": "toggle"
888
+ }
889
+ ],
890
+ "tool_call": true,
891
+ "structured_output": true,
892
+ "temperature": true,
893
+ "knowledge": "2025-05",
894
+ "release_date": "2026-07-31",
895
+ "last_updated": "2026-07-31",
896
+ "modalities": {
897
+ "input": [
898
+ "text"
899
+ ],
900
+ "output": [
901
+ "text"
902
+ ]
903
+ },
904
+ "open_weights": true,
905
+ "limit": {
906
+ "context": 1048576,
907
+ "output": 384000
908
+ },
909
+ "cost": {
910
+ "input": 0.08,
911
+ "output": 0.18,
912
+ "cache_read": 0.016
913
+ }
914
+ },
915
+ "deepseek-ai/DeepSeek-V3.2": {
916
+ "id": "deepseek-ai/DeepSeek-V3.2",
917
+ "name": "DeepSeek-V3.2",
918
+ "description": "DeepSeek chat model for instruction following, coding, and analysis",
919
+ "attachment": false,
920
+ "reasoning": true,
921
+ "reasoning_options": [
922
+ {
923
+ "type": "toggle"
924
+ }
925
+ ],
926
+ "tool_call": true,
927
+ "interleaved": {
928
+ "field": "reasoning_content"
929
+ },
930
+ "structured_output": true,
931
+ "temperature": true,
932
+ "knowledge": "2024-12",
933
+ "release_date": "2025-12-02",
934
+ "last_updated": "2025-12-02",
935
+ "modalities": {
936
+ "input": [
937
+ "text"
938
+ ],
939
+ "output": [
940
+ "text"
941
+ ]
942
+ },
943
+ "open_weights": false,
944
+ "limit": {
945
+ "context": 163840,
946
+ "output": 64000
947
+ },
948
+ "cost": {
949
+ "input": 0.26,
950
+ "output": 0.38,
951
+ "cache_read": 0.13
952
+ }
953
+ },
954
+ "deepseek-ai/DeepSeek-R1-0528": {
955
+ "id": "deepseek-ai/DeepSeek-R1-0528",
956
+ "name": "DeepSeek-R1-0528",
957
+ "description": "DeepSeek reasoning model for multi-step analysis, math, coding, and tools",
958
+ "attachment": false,
959
+ "reasoning": true,
960
+ "reasoning_options": [],
961
+ "tool_call": true,
962
+ "interleaved": {
963
+ "field": "reasoning_content"
964
+ },
965
+ "structured_output": true,
966
+ "temperature": true,
967
+ "knowledge": "2024-07",
968
+ "release_date": "2025-05-28",
969
+ "last_updated": "2025-05-28",
970
+ "modalities": {
971
+ "input": [
972
+ "text"
973
+ ],
974
+ "output": [
975
+ "text"
976
+ ]
977
+ },
978
+ "open_weights": false,
979
+ "limit": {
980
+ "context": 163840,
981
+ "output": 64000
982
+ },
983
+ "cost": {
984
+ "input": 0.5,
985
+ "output": 2.15,
986
+ "cache_read": 0.35
987
+ }
988
+ },
989
+ "deepseek-ai/DeepSeek-V3.1": {
990
+ "id": "deepseek-ai/DeepSeek-V3.1",
991
+ "name": "DeepSeek-V3.1",
992
+ "description": "Hybrid-reasoning DeepSeek model with thinking and non-thinking modes",
993
+ "family": "deepseek",
994
+ "attachment": false,
995
+ "reasoning": true,
996
+ "reasoning_options": [
997
+ {
998
+ "type": "toggle"
999
+ }
1000
+ ],
1001
+ "tool_call": true,
1002
+ "structured_output": true,
1003
+ "temperature": true,
1004
+ "release_date": "2025-08-21",
1005
+ "last_updated": "2025-08-21",
1006
+ "modalities": {
1007
+ "input": [
1008
+ "text"
1009
+ ],
1010
+ "output": [
1011
+ "text"
1012
+ ]
1013
+ },
1014
+ "open_weights": true,
1015
+ "limit": {
1016
+ "context": 163840,
1017
+ "output": 8192
1018
+ },
1019
+ "cost": {
1020
+ "input": 0.25,
1021
+ "output": 0.95,
1022
+ "cache_read": 0.13
1023
+ }
1024
+ },
1025
+ "deepseek-ai/DeepSeek-V3-0324": {
1026
+ "id": "deepseek-ai/DeepSeek-V3-0324",
1027
+ "name": "DeepSeek V3 0324",
1028
+ "description": "March 2025 checkpoint of DeepSeek-V3 with improved reasoning and coding",
1029
+ "family": "deepseek",
1030
+ "attachment": false,
1031
+ "reasoning": false,
1032
+ "tool_call": true,
1033
+ "structured_output": true,
1034
+ "temperature": true,
1035
+ "release_date": "2025-03-24",
1036
+ "last_updated": "2025-03-24",
1037
+ "modalities": {
1038
+ "input": [
1039
+ "text"
1040
+ ],
1041
+ "output": [
1042
+ "text"
1043
+ ]
1044
+ },
1045
+ "open_weights": true,
1046
+ "limit": {
1047
+ "context": 163840,
1048
+ "output": 163840
1049
+ },
1050
+ "cost": {
1051
+ "input": 0.24,
1052
+ "output": 0.9,
1053
+ "cache_read": 0.135
1054
+ }
1055
+ },
1056
+ "deepseek-ai/DeepSeek-V3": {
1057
+ "id": "deepseek-ai/DeepSeek-V3",
1058
+ "name": "DeepSeek-V3",
1059
+ "description": "Open DeepSeek MoE chat model for coding, math, and general reasoning",
1060
+ "family": "deepseek",
1061
+ "attachment": false,
1062
+ "reasoning": false,
1063
+ "tool_call": true,
1064
+ "structured_output": true,
1065
+ "temperature": true,
1066
+ "release_date": "2024-12-26",
1067
+ "last_updated": "2024-12-26",
1068
+ "modalities": {
1069
+ "input": [
1070
+ "text"
1071
+ ],
1072
+ "output": [
1073
+ "text"
1074
+ ]
1075
+ },
1076
+ "open_weights": true,
1077
+ "limit": {
1078
+ "context": 163840,
1079
+ "output": 8192
625
1080
  },
626
1081
  "cost": {
627
- "input": 0.1,
628
- "output": 0.15
1082
+ "input": 0.32,
1083
+ "output": 0.89
629
1084
  }
630
1085
  },
631
- "Qwen/Qwen3-235B-A22B-Instruct-2507": {
632
- "id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
633
- "name": "Qwen3 235B-A22B Instruct 2507",
634
- "description": "Updated large open Qwen3 MoE instruct model for multilingual chat, coding, and tool use",
635
- "family": "qwen",
1086
+ "deepseek-ai/DeepSeek-V4-Flash": {
1087
+ "id": "deepseek-ai/DeepSeek-V4-Flash",
1088
+ "name": "DeepSeek V4 Flash",
1089
+ "description": "Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work",
1090
+ "family": "deepseek-flash",
636
1091
  "attachment": false,
637
- "reasoning": false,
1092
+ "reasoning": true,
1093
+ "reasoning_options": [
1094
+ {
1095
+ "type": "toggle"
1096
+ },
1097
+ {
1098
+ "type": "effort",
1099
+ "values": [
1100
+ "low",
1101
+ "medium",
1102
+ "high",
1103
+ "xhigh"
1104
+ ]
1105
+ }
1106
+ ],
638
1107
  "tool_call": true,
1108
+ "interleaved": {
1109
+ "field": "reasoning_content"
1110
+ },
639
1111
  "structured_output": true,
640
1112
  "temperature": true,
641
- "release_date": "2025-07-21",
642
- "last_updated": "2025-07-21",
1113
+ "knowledge": "2025-05",
1114
+ "release_date": "2026-04-24",
1115
+ "last_updated": "2026-04-24",
643
1116
  "modalities": {
644
1117
  "input": [
645
1118
  "text"
@@ -650,33 +1123,32 @@
650
1123
  },
651
1124
  "open_weights": true,
652
1125
  "limit": {
653
- "context": 262144,
1126
+ "context": 1048576,
654
1127
  "output": 16384
655
1128
  },
656
1129
  "cost": {
657
1130
  "input": 0.09,
658
- "output": 0.55
1131
+ "output": 0.18,
1132
+ "cache_read": 0.018
659
1133
  }
660
1134
  },
661
- "Qwen/Qwen3.5-122B-A10B": {
662
- "id": "Qwen/Qwen3.5-122B-A10B",
663
- "name": "Qwen3.5 122B-A10B",
664
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
665
- "family": "qwen",
1135
+ "stepfun-ai/Step-3.7-Flash": {
1136
+ "id": "stepfun-ai/Step-3.7-Flash",
1137
+ "name": "Step 3.7 Flash",
1138
+ "description": "Newer StepFun flash model for faster agents, coding, and multimodal prompts",
666
1139
  "attachment": true,
667
1140
  "reasoning": true,
668
1141
  "reasoning_options": [],
669
1142
  "tool_call": true,
670
- "structured_output": true,
671
1143
  "temperature": true,
672
- "release_date": "2026-02-23",
673
- "last_updated": "2026-02-23",
1144
+ "knowledge": "2026-03-01",
1145
+ "release_date": "2026-05-29",
1146
+ "last_updated": "2026-05-29",
674
1147
  "modalities": {
675
1148
  "input": [
676
1149
  "text",
677
1150
  "image",
678
- "video",
679
- "audio"
1151
+ "video"
680
1152
  ],
681
1153
  "output": [
682
1154
  "text"
@@ -685,25 +1157,26 @@
685
1157
  "open_weights": true,
686
1158
  "limit": {
687
1159
  "context": 262144,
688
- "output": 65536
1160
+ "output": 256000
689
1161
  },
690
1162
  "cost": {
691
- "input": 0.29,
692
- "output": 2.4
1163
+ "input": 0.2,
1164
+ "output": 1.15,
1165
+ "cache_read": 0.04
693
1166
  }
694
1167
  },
695
- "Qwen/Qwen3.7-Max": {
696
- "id": "Qwen/Qwen3.7-Max",
697
- "name": "Qwen3.7 Max",
698
- "description": "Qwen frontier model tuned for agent frameworks, coding assistants, and long tasks",
1168
+ "Qwen/Qwen3-235B-A22B-Instruct-2507": {
1169
+ "id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
1170
+ "name": "Qwen3 235B-A22B Instruct 2507",
1171
+ "description": "Updated large open Qwen3 MoE instruct model for multilingual chat, coding, and tool use",
699
1172
  "family": "qwen",
700
1173
  "attachment": false,
701
1174
  "reasoning": false,
702
1175
  "tool_call": true,
703
1176
  "structured_output": true,
704
1177
  "temperature": true,
705
- "release_date": "2026-05-21",
706
- "last_updated": "2026-05-21",
1178
+ "release_date": "2025-07-21",
1179
+ "last_updated": "2025-07-21",
707
1180
  "modalities": {
708
1181
  "input": [
709
1182
  "text"
@@ -712,70 +1185,14 @@
712
1185
  "text"
713
1186
  ]
714
1187
  },
715
- "open_weights": false,
716
- "limit": {
717
- "context": 256000,
718
- "output": 65536
719
- },
720
- "cost": {
721
- "input": 2.5,
722
- "output": 7.5,
723
- "cache_read": 0.5,
724
- "tiers": [
725
- {
726
- "input": 5,
727
- "output": 15,
728
- "cache_read": 1,
729
- "tier": {
730
- "type": "context",
731
- "size": 32000
732
- }
733
- },
734
- {
735
- "input": 6.25,
736
- "output": 18.5,
737
- "cache_read": 1.25,
738
- "tier": {
739
- "type": "context",
740
- "size": 128000
741
- }
742
- }
743
- ]
744
- }
745
- },
746
- "Qwen/Qwen3.5-35B-A3B": {
747
- "id": "Qwen/Qwen3.5-35B-A3B",
748
- "name": "Qwen 3.5 35B A3B",
749
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
750
- "family": "qwen",
751
- "attachment": true,
752
- "reasoning": true,
753
- "reasoning_options": [],
754
- "tool_call": true,
755
- "structured_output": true,
756
- "temperature": true,
757
- "knowledge": "2025-01",
758
- "release_date": "2026-02-01",
759
- "last_updated": "2026-04-20",
760
- "modalities": {
761
- "input": [
762
- "text",
763
- "image",
764
- "video"
765
- ],
766
- "output": [
767
- "text"
768
- ]
769
- },
770
1188
  "open_weights": true,
771
1189
  "limit": {
772
1190
  "context": 262144,
773
- "output": 81920
1191
+ "output": 16384
774
1192
  },
775
1193
  "cost": {
776
- "input": 0.14,
777
- "output": 1,
778
- "cache_read": 0.05
1194
+ "input": 0.09,
1195
+ "output": 0.55
779
1196
  }
780
1197
  },
781
1198
  "Qwen/Qwen3.6-27B": {
@@ -812,10 +1229,10 @@
812
1229
  "output": 3.2
813
1230
  }
814
1231
  },
815
- "Qwen/Qwen3.5-397B-A17B": {
816
- "id": "Qwen/Qwen3.5-397B-A17B",
817
- "name": "Qwen 3.5 397B A17B",
818
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1232
+ "Qwen/Qwen3.5-9B": {
1233
+ "id": "Qwen/Qwen3.5-9B",
1234
+ "name": "Qwen3.5 9B",
1235
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
819
1236
  "family": "qwen",
820
1237
  "attachment": true,
821
1238
  "reasoning": true,
@@ -823,9 +1240,8 @@
823
1240
  "tool_call": true,
824
1241
  "structured_output": true,
825
1242
  "temperature": true,
826
- "knowledge": "2025-01",
827
- "release_date": "2026-02-01",
828
- "last_updated": "2026-04-20",
1243
+ "release_date": "2026-02-23",
1244
+ "last_updated": "2026-02-23",
829
1245
  "modalities": {
830
1246
  "input": [
831
1247
  "text",
@@ -839,32 +1255,29 @@
839
1255
  "open_weights": true,
840
1256
  "limit": {
841
1257
  "context": 262144,
842
- "output": 81920
1258
+ "output": 65536
843
1259
  },
844
1260
  "cost": {
845
- "input": 0.45,
846
- "output": 3,
847
- "cache_read": 0.22
1261
+ "input": 0.1,
1262
+ "output": 0.15
848
1263
  }
849
1264
  },
850
- "Qwen/Qwen3.6-35B-A3B": {
851
- "id": "Qwen/Qwen3.6-35B-A3B",
852
- "name": "Qwen3.6 35B A3B",
853
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1265
+ "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo": {
1266
+ "id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo",
1267
+ "name": "Qwen3 Coder 480B A35B Instruct Turbo",
1268
+ "description": "Qwen coding model for software agents, repository edits, and code reasoning",
854
1269
  "family": "qwen",
855
- "attachment": true,
856
- "reasoning": true,
857
- "reasoning_options": [],
1270
+ "attachment": false,
1271
+ "reasoning": false,
858
1272
  "tool_call": true,
859
1273
  "structured_output": true,
860
1274
  "temperature": true,
861
- "release_date": "2026-04-01",
862
- "last_updated": "2026-04-01",
1275
+ "knowledge": "2025-04",
1276
+ "release_date": "2025-07-23",
1277
+ "last_updated": "2025-07-23",
863
1278
  "modalities": {
864
1279
  "input": [
865
- "text",
866
- "image",
867
- "video"
1280
+ "text"
868
1281
  ],
869
1282
  "output": [
870
1283
  "text"
@@ -873,45 +1286,46 @@
873
1286
  "open_weights": true,
874
1287
  "limit": {
875
1288
  "context": 262144,
876
- "output": 81920
1289
+ "output": 66536
877
1290
  },
878
1291
  "cost": {
879
- "input": 0.1,
880
- "output": 0.95
1292
+ "input": 0.3,
1293
+ "output": 1,
1294
+ "cache_read": 0.1
881
1295
  }
882
1296
  },
883
- "Qwen/Qwen3.8-Max": {
884
- "id": "Qwen/Qwen3.8-Max",
885
- "name": "Qwen3.8 Max",
886
- "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
1297
+ "Qwen/Qwen3.5-122B-A10B": {
1298
+ "id": "Qwen/Qwen3.5-122B-A10B",
1299
+ "name": "Qwen3.5 122B-A10B",
1300
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
887
1301
  "family": "qwen",
888
1302
  "attachment": true,
889
- "reasoning": false,
1303
+ "reasoning": true,
1304
+ "reasoning_options": [],
890
1305
  "tool_call": true,
891
1306
  "structured_output": true,
892
1307
  "temperature": true,
893
- "release_date": "2026-08-03",
894
- "last_updated": "2026-08-03",
1308
+ "release_date": "2026-02-23",
1309
+ "last_updated": "2026-02-23",
895
1310
  "modalities": {
896
1311
  "input": [
897
1312
  "text",
898
1313
  "image",
899
1314
  "video",
900
- "pdf"
1315
+ "audio"
901
1316
  ],
902
1317
  "output": [
903
1318
  "text"
904
1319
  ]
905
1320
  },
906
- "open_weights": false,
1321
+ "open_weights": true,
907
1322
  "limit": {
908
- "context": 256000,
909
- "output": 131072
1323
+ "context": 262144,
1324
+ "output": 65536
910
1325
  },
911
1326
  "cost": {
912
- "input": 1.65,
913
- "output": 4.951,
914
- "cache_read": 0.206
1327
+ "input": 0.29,
1328
+ "output": 2.4
915
1329
  }
916
1330
  },
917
1331
  "Qwen/Qwen3-32B": {
@@ -927,38 +1341,7 @@
927
1341
  "temperature": true,
928
1342
  "knowledge": "2025-04",
929
1343
  "release_date": "2025-04",
930
- "last_updated": "2025-04",
931
- "modalities": {
932
- "input": [
933
- "text"
934
- ],
935
- "output": [
936
- "text"
937
- ]
938
- },
939
- "open_weights": true,
940
- "limit": {
941
- "context": 40960,
942
- "output": 16384
943
- },
944
- "cost": {
945
- "input": 0.08,
946
- "output": 0.28
947
- }
948
- },
949
- "Qwen/Qwen3-Next-80B-A3B-Instruct": {
950
- "id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
951
- "name": "Qwen3-Next 80B-A3B Instruct",
952
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
953
- "family": "qwen",
954
- "attachment": false,
955
- "reasoning": false,
956
- "tool_call": true,
957
- "structured_output": true,
958
- "temperature": true,
959
- "knowledge": "2025-04",
960
- "release_date": "2025-09",
961
- "last_updated": "2025-09",
1344
+ "last_updated": "2025-04",
962
1345
  "modalities": {
963
1346
  "input": [
964
1347
  "text"
@@ -969,59 +1352,60 @@
969
1352
  },
970
1353
  "open_weights": true,
971
1354
  "limit": {
972
- "context": 262144,
973
- "output": 32768
1355
+ "context": 40960,
1356
+ "output": 16384
974
1357
  },
975
1358
  "cost": {
976
- "input": 0.09,
977
- "output": 1.1
1359
+ "input": 0.08,
1360
+ "output": 0.28
978
1361
  }
979
1362
  },
980
- "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo": {
981
- "id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo",
982
- "name": "Qwen3 Coder 480B A35B Instruct Turbo",
983
- "description": "Qwen coding model for software agents, repository edits, and code reasoning",
1363
+ "Qwen/Qwen3.8-Max": {
1364
+ "id": "Qwen/Qwen3.8-Max",
1365
+ "name": "Qwen3.8 Max",
1366
+ "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
984
1367
  "family": "qwen",
985
- "attachment": false,
1368
+ "attachment": true,
986
1369
  "reasoning": false,
987
1370
  "tool_call": true,
988
1371
  "structured_output": true,
989
1372
  "temperature": true,
990
- "knowledge": "2025-04",
991
- "release_date": "2025-07-23",
992
- "last_updated": "2025-07-23",
1373
+ "release_date": "2026-08-03",
1374
+ "last_updated": "2026-08-03",
993
1375
  "modalities": {
994
1376
  "input": [
995
- "text"
1377
+ "text",
1378
+ "image",
1379
+ "video",
1380
+ "pdf"
996
1381
  ],
997
1382
  "output": [
998
1383
  "text"
999
1384
  ]
1000
1385
  },
1001
- "open_weights": true,
1386
+ "open_weights": false,
1002
1387
  "limit": {
1003
- "context": 262144,
1004
- "output": 66536
1388
+ "context": 256000,
1389
+ "output": 131072
1005
1390
  },
1006
1391
  "cost": {
1007
- "input": 0.3,
1008
- "output": 1,
1009
- "cache_read": 0.1
1392
+ "input": 1.65,
1393
+ "output": 4.951,
1394
+ "cache_read": 0.206
1010
1395
  }
1011
1396
  },
1012
- "Qwen/Qwen3-Max": {
1013
- "id": "Qwen/Qwen3-Max",
1014
- "name": "Qwen3 Max",
1015
- "description": "Flagship Qwen3 model for coding agents, complex reasoning, and tool use",
1397
+ "Qwen/Qwen3.7-Max": {
1398
+ "id": "Qwen/Qwen3.7-Max",
1399
+ "name": "Qwen3.7 Max",
1400
+ "description": "Qwen frontier model tuned for agent frameworks, coding assistants, and long tasks",
1016
1401
  "family": "qwen",
1017
1402
  "attachment": false,
1018
1403
  "reasoning": false,
1019
1404
  "tool_call": true,
1020
1405
  "structured_output": true,
1021
1406
  "temperature": true,
1022
- "knowledge": "2025-04",
1023
- "release_date": "2025-09-23",
1024
- "last_updated": "2025-09-23",
1407
+ "release_date": "2026-05-21",
1408
+ "last_updated": "2026-05-21",
1025
1409
  "modalities": {
1026
1410
  "input": [
1027
1411
  "text"
@@ -1036,23 +1420,23 @@
1036
1420
  "output": 65536
1037
1421
  },
1038
1422
  "cost": {
1039
- "input": 1.2,
1040
- "output": 6,
1041
- "cache_read": 0.24,
1423
+ "input": 2.5,
1424
+ "output": 7.5,
1425
+ "cache_read": 0.5,
1042
1426
  "tiers": [
1043
1427
  {
1044
- "input": 2.4,
1045
- "output": 12,
1046
- "cache_read": 0.48,
1428
+ "input": 5,
1429
+ "output": 15,
1430
+ "cache_read": 1,
1047
1431
  "tier": {
1048
1432
  "type": "context",
1049
1433
  "size": 32000
1050
1434
  }
1051
1435
  },
1052
1436
  {
1053
- "input": 3,
1054
- "output": 15,
1055
- "cache_read": 0.6,
1437
+ "input": 6.25,
1438
+ "output": 18.5,
1439
+ "cache_read": 1.25,
1056
1440
  "tier": {
1057
1441
  "type": "context",
1058
1442
  "size": 128000
@@ -1061,56 +1445,25 @@
1061
1445
  ]
1062
1446
  }
1063
1447
  },
1064
- "MiniMaxAI/MiniMax-M2.7": {
1065
- "id": "MiniMaxAI/MiniMax-M2.7",
1066
- "name": "MiniMax-M2.7",
1067
- "description": "Open MiniMax flagship for coding agents, office automation, and complex environments",
1068
- "family": "minimax",
1069
- "attachment": false,
1070
- "reasoning": true,
1071
- "reasoning_options": [],
1072
- "tool_call": true,
1073
- "temperature": true,
1074
- "release_date": "2026-03-18",
1075
- "last_updated": "2026-03-18",
1076
- "modalities": {
1077
- "input": [
1078
- "text"
1079
- ],
1080
- "output": [
1081
- "text"
1082
- ]
1083
- },
1084
- "open_weights": true,
1085
- "limit": {
1086
- "context": 196608,
1087
- "output": 131072
1088
- },
1089
- "cost": {
1090
- "input": 0.25,
1091
- "output": 1,
1092
- "cache_read": 0.05
1093
- }
1094
- },
1095
- "MiniMaxAI/MiniMax-M2.5": {
1096
- "id": "MiniMaxAI/MiniMax-M2.5",
1097
- "name": "MiniMax M2.5",
1098
- "description": "MiniMax model for chat, coding, office work, and agentic tasks",
1099
- "family": "minimax",
1100
- "attachment": false,
1448
+ "Qwen/Qwen3.5-27B": {
1449
+ "id": "Qwen/Qwen3.5-27B",
1450
+ "name": "Qwen3.5 27B",
1451
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1452
+ "family": "qwen",
1453
+ "attachment": true,
1101
1454
  "reasoning": true,
1102
1455
  "reasoning_options": [],
1103
1456
  "tool_call": true,
1104
- "interleaved": {
1105
- "field": "reasoning_content"
1106
- },
1457
+ "structured_output": true,
1107
1458
  "temperature": true,
1108
- "knowledge": "2025-06",
1109
- "release_date": "2026-02-12",
1110
- "last_updated": "2026-02-12",
1459
+ "release_date": "2026-02-23",
1460
+ "last_updated": "2026-02-23",
1111
1461
  "modalities": {
1112
1462
  "input": [
1113
- "text"
1463
+ "text",
1464
+ "image",
1465
+ "video",
1466
+ "audio"
1114
1467
  ],
1115
1468
  "output": [
1116
1469
  "text"
@@ -1118,29 +1471,28 @@
1118
1471
  },
1119
1472
  "open_weights": true,
1120
1473
  "limit": {
1121
- "context": 196608,
1122
- "output": 131072
1474
+ "context": 262144,
1475
+ "output": 65536
1123
1476
  },
1124
- "status": "deprecated",
1125
1477
  "cost": {
1126
- "input": 0.15,
1127
- "output": 1.15,
1128
- "cache_read": 0.03
1478
+ "input": 0.26,
1479
+ "output": 2.6
1129
1480
  }
1130
1481
  },
1131
- "MiniMaxAI/MiniMax-M3": {
1132
- "id": "MiniMaxAI/MiniMax-M3",
1133
- "name": "MiniMax-M3",
1134
- "description": "MiniMax multimodal model for long-context coding, perception, and agent planning",
1135
- "family": "minimax",
1482
+ "Qwen/Qwen3.5-35B-A3B": {
1483
+ "id": "Qwen/Qwen3.5-35B-A3B",
1484
+ "name": "Qwen 3.5 35B A3B",
1485
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1486
+ "family": "qwen",
1136
1487
  "attachment": true,
1137
1488
  "reasoning": true,
1138
1489
  "reasoning_options": [],
1139
1490
  "tool_call": true,
1140
1491
  "structured_output": true,
1141
1492
  "temperature": true,
1142
- "release_date": "2026-06-01",
1143
- "last_updated": "2026-06-01",
1493
+ "knowledge": "2025-01",
1494
+ "release_date": "2026-02-01",
1495
+ "last_updated": "2026-04-20",
1144
1496
  "modalities": {
1145
1497
  "input": [
1146
1498
  "text",
@@ -1153,27 +1505,28 @@
1153
1505
  },
1154
1506
  "open_weights": true,
1155
1507
  "limit": {
1156
- "context": 524288,
1157
- "output": 128000
1508
+ "context": 262144,
1509
+ "output": 81920
1158
1510
  },
1159
1511
  "cost": {
1160
- "input": 0.3,
1161
- "output": 1.2,
1162
- "cache_read": 0.06
1512
+ "input": 0.14,
1513
+ "output": 1,
1514
+ "cache_read": 0.05
1163
1515
  }
1164
1516
  },
1165
- "deepseek-ai/DeepSeek-V3": {
1166
- "id": "deepseek-ai/DeepSeek-V3",
1167
- "name": "DeepSeek-V3",
1168
- "description": "Open DeepSeek MoE chat model for coding, math, and general reasoning",
1169
- "family": "deepseek",
1517
+ "Qwen/Qwen3-Max": {
1518
+ "id": "Qwen/Qwen3-Max",
1519
+ "name": "Qwen3 Max",
1520
+ "description": "Flagship Qwen3 model for coding agents, complex reasoning, and tool use",
1521
+ "family": "qwen",
1170
1522
  "attachment": false,
1171
1523
  "reasoning": false,
1172
1524
  "tool_call": true,
1173
1525
  "structured_output": true,
1174
1526
  "temperature": true,
1175
- "release_date": "2024-12-26",
1176
- "last_updated": "2024-12-26",
1527
+ "knowledge": "2025-04",
1528
+ "release_date": "2025-09-23",
1529
+ "last_updated": "2025-09-23",
1177
1530
  "modalities": {
1178
1531
  "input": [
1179
1532
  "text"
@@ -1182,69 +1535,50 @@
1182
1535
  "text"
1183
1536
  ]
1184
1537
  },
1185
- "open_weights": true,
1538
+ "open_weights": false,
1186
1539
  "limit": {
1187
- "context": 163840,
1188
- "output": 8192
1540
+ "context": 256000,
1541
+ "output": 65536
1189
1542
  },
1190
1543
  "cost": {
1191
- "input": 0.32,
1192
- "output": 0.89
1193
- }
1194
- },
1195
- "deepseek-ai/DeepSeek-V4-Flash-0731": {
1196
- "id": "deepseek-ai/DeepSeek-V4-Flash-0731",
1197
- "name": "DeepSeek V4 Flash 0731",
1198
- "description": "Official DeepSeek V4 Flash release with enhanced agentic capabilities and integrated DSpark speculative decoding",
1199
- "family": "deepseek-flash",
1200
- "attachment": false,
1201
- "reasoning": true,
1202
- "reasoning_options": [
1203
- {
1204
- "type": "toggle"
1205
- }
1206
- ],
1207
- "tool_call": true,
1208
- "structured_output": true,
1209
- "temperature": true,
1210
- "knowledge": "2025-05",
1211
- "release_date": "2026-07-31",
1212
- "last_updated": "2026-07-31",
1213
- "modalities": {
1214
- "input": [
1215
- "text"
1216
- ],
1217
- "output": [
1218
- "text"
1544
+ "input": 1.2,
1545
+ "output": 6,
1546
+ "cache_read": 0.24,
1547
+ "tiers": [
1548
+ {
1549
+ "input": 2.4,
1550
+ "output": 12,
1551
+ "cache_read": 0.48,
1552
+ "tier": {
1553
+ "type": "context",
1554
+ "size": 32000
1555
+ }
1556
+ },
1557
+ {
1558
+ "input": 3,
1559
+ "output": 15,
1560
+ "cache_read": 0.6,
1561
+ "tier": {
1562
+ "type": "context",
1563
+ "size": 128000
1564
+ }
1565
+ }
1219
1566
  ]
1220
- },
1221
- "open_weights": true,
1222
- "limit": {
1223
- "context": 1048576,
1224
- "output": 384000
1225
- },
1226
- "cost": {
1227
- "input": 0.09,
1228
- "output": 0.18,
1229
- "cache_read": 0.018
1230
1567
  }
1231
1568
  },
1232
- "deepseek-ai/DeepSeek-R1-0528": {
1233
- "id": "deepseek-ai/DeepSeek-R1-0528",
1234
- "name": "DeepSeek-R1-0528",
1235
- "description": "DeepSeek reasoning model for multi-step analysis, math, coding, and tools",
1569
+ "Qwen/Qwen3-30B-A3B": {
1570
+ "id": "Qwen/Qwen3-30B-A3B",
1571
+ "name": "Qwen3 30B A3B",
1572
+ "description": "Sparse MoE Qwen model with 3B active parameters for efficient chat and reasoning",
1573
+ "family": "qwen",
1236
1574
  "attachment": false,
1237
1575
  "reasoning": true,
1238
1576
  "reasoning_options": [],
1239
1577
  "tool_call": true,
1240
- "interleaved": {
1241
- "field": "reasoning_content"
1242
- },
1243
1578
  "structured_output": true,
1244
- "temperature": true,
1245
- "knowledge": "2024-07",
1246
- "release_date": "2025-05-28",
1247
- "last_updated": "2025-05-28",
1579
+ "temperature": true,
1580
+ "release_date": "2025-04-28",
1581
+ "last_updated": "2025-04-28",
1248
1582
  "modalities": {
1249
1583
  "input": [
1250
1584
  "text"
@@ -1253,76 +1587,69 @@
1253
1587
  "text"
1254
1588
  ]
1255
1589
  },
1256
- "open_weights": false,
1590
+ "open_weights": true,
1257
1591
  "limit": {
1258
- "context": 163840,
1259
- "output": 64000
1592
+ "context": 40960,
1593
+ "output": 16384
1260
1594
  },
1261
1595
  "cost": {
1262
- "input": 0.5,
1263
- "output": 2.15,
1264
- "cache_read": 0.35
1596
+ "input": 0.12,
1597
+ "output": 0.5
1265
1598
  }
1266
1599
  },
1267
- "deepseek-ai/DeepSeek-V3.2": {
1268
- "id": "deepseek-ai/DeepSeek-V3.2",
1269
- "name": "DeepSeek-V3.2",
1270
- "description": "DeepSeek chat model for instruction following, coding, and analysis",
1271
- "attachment": false,
1600
+ "Qwen/Qwen3.5-397B-A17B": {
1601
+ "id": "Qwen/Qwen3.5-397B-A17B",
1602
+ "name": "Qwen 3.5 397B A17B",
1603
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1604
+ "family": "qwen",
1605
+ "attachment": true,
1272
1606
  "reasoning": true,
1273
- "reasoning_options": [
1274
- {
1275
- "type": "toggle"
1276
- }
1277
- ],
1607
+ "reasoning_options": [],
1278
1608
  "tool_call": true,
1279
- "interleaved": {
1280
- "field": "reasoning_content"
1281
- },
1282
1609
  "structured_output": true,
1283
1610
  "temperature": true,
1284
- "knowledge": "2024-12",
1285
- "release_date": "2025-12-02",
1286
- "last_updated": "2025-12-02",
1611
+ "knowledge": "2025-01",
1612
+ "release_date": "2026-02-01",
1613
+ "last_updated": "2026-04-20",
1287
1614
  "modalities": {
1288
1615
  "input": [
1289
- "text"
1616
+ "text",
1617
+ "image",
1618
+ "video"
1290
1619
  ],
1291
1620
  "output": [
1292
1621
  "text"
1293
1622
  ]
1294
1623
  },
1295
- "open_weights": false,
1624
+ "open_weights": true,
1296
1625
  "limit": {
1297
- "context": 163840,
1298
- "output": 64000
1626
+ "context": 262144,
1627
+ "output": 81920
1299
1628
  },
1300
1629
  "cost": {
1301
- "input": 0.26,
1302
- "output": 0.38,
1303
- "cache_read": 0.13
1630
+ "input": 0.45,
1631
+ "output": 3,
1632
+ "cache_read": 0.22
1304
1633
  }
1305
1634
  },
1306
- "deepseek-ai/DeepSeek-V3.1": {
1307
- "id": "deepseek-ai/DeepSeek-V3.1",
1308
- "name": "DeepSeek-V3.1",
1309
- "description": "Hybrid-reasoning DeepSeek model with thinking and non-thinking modes",
1310
- "family": "deepseek",
1311
- "attachment": false,
1635
+ "Qwen/Qwen3.6-35B-A3B": {
1636
+ "id": "Qwen/Qwen3.6-35B-A3B",
1637
+ "name": "Qwen3.6 35B A3B",
1638
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1639
+ "family": "qwen",
1640
+ "attachment": true,
1312
1641
  "reasoning": true,
1313
- "reasoning_options": [
1314
- {
1315
- "type": "toggle"
1316
- }
1317
- ],
1642
+ "reasoning_options": [],
1318
1643
  "tool_call": true,
1319
1644
  "structured_output": true,
1320
1645
  "temperature": true,
1321
- "release_date": "2025-08-21",
1322
- "last_updated": "2025-08-21",
1646
+ "release_date": "2026-04-01",
1647
+ "last_updated": "2026-04-01",
1323
1648
  "modalities": {
1324
1649
  "input": [
1325
- "text"
1650
+ "text",
1651
+ "image",
1652
+ "video"
1326
1653
  ],
1327
1654
  "output": [
1328
1655
  "text"
@@ -1330,45 +1657,27 @@
1330
1657
  },
1331
1658
  "open_weights": true,
1332
1659
  "limit": {
1333
- "context": 163840,
1334
- "output": 8192
1660
+ "context": 262144,
1661
+ "output": 81920
1335
1662
  },
1336
1663
  "cost": {
1337
- "input": 0.25,
1338
- "output": 0.95,
1339
- "cache_read": 0.13
1664
+ "input": 0.1,
1665
+ "output": 0.95
1340
1666
  }
1341
1667
  },
1342
- "deepseek-ai/DeepSeek-V4-Flash": {
1343
- "id": "deepseek-ai/DeepSeek-V4-Flash",
1344
- "name": "DeepSeek V4 Flash",
1345
- "description": "Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work",
1346
- "family": "deepseek-flash",
1668
+ "Qwen/Qwen3-Next-80B-A3B-Instruct": {
1669
+ "id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
1670
+ "name": "Qwen3-Next 80B-A3B Instruct",
1671
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1672
+ "family": "qwen",
1347
1673
  "attachment": false,
1348
- "reasoning": true,
1349
- "reasoning_options": [
1350
- {
1351
- "type": "toggle"
1352
- },
1353
- {
1354
- "type": "effort",
1355
- "values": [
1356
- "low",
1357
- "medium",
1358
- "high",
1359
- "xhigh"
1360
- ]
1361
- }
1362
- ],
1674
+ "reasoning": false,
1363
1675
  "tool_call": true,
1364
- "interleaved": {
1365
- "field": "reasoning_content"
1366
- },
1367
1676
  "structured_output": true,
1368
1677
  "temperature": true,
1369
- "knowledge": "2025-05",
1370
- "release_date": "2026-04-24",
1371
- "last_updated": "2026-04-24",
1678
+ "knowledge": "2025-04",
1679
+ "release_date": "2025-09",
1680
+ "last_updated": "2025-09",
1372
1681
  "modalities": {
1373
1682
  "input": [
1374
1683
  "text"
@@ -1379,34 +1688,24 @@
1379
1688
  },
1380
1689
  "open_weights": true,
1381
1690
  "limit": {
1382
- "context": 1048576,
1383
- "output": 16384
1691
+ "context": 262144,
1692
+ "output": 32768
1384
1693
  },
1385
1694
  "cost": {
1386
1695
  "input": 0.09,
1387
- "output": 0.18,
1388
- "cache_read": 0.018
1696
+ "output": 1.1
1389
1697
  }
1390
1698
  },
1391
- "deepseek-ai/DeepSeek-V4-Pro": {
1392
- "id": "deepseek-ai/DeepSeek-V4-Pro",
1393
- "name": "DeepSeek V4 Pro",
1394
- "description": "Open MoE flagship with million-token context for coding and long agent runs",
1395
- "family": "deepseek-thinking",
1699
+ "zai-org/GLM-5.1": {
1700
+ "id": "zai-org/GLM-5.1",
1701
+ "name": "GLM-5.1",
1702
+ "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
1703
+ "family": "glm",
1396
1704
  "attachment": false,
1397
1705
  "reasoning": true,
1398
1706
  "reasoning_options": [
1399
1707
  {
1400
1708
  "type": "toggle"
1401
- },
1402
- {
1403
- "type": "effort",
1404
- "values": [
1405
- "low",
1406
- "medium",
1407
- "high",
1408
- "xhigh"
1409
- ]
1410
1709
  }
1411
1710
  ],
1412
1711
  "tool_call": true,
@@ -1415,9 +1714,9 @@
1415
1714
  },
1416
1715
  "structured_output": true,
1417
1716
  "temperature": true,
1418
- "knowledge": "2025-05",
1419
- "release_date": "2026-04-24",
1420
- "last_updated": "2026-04-24",
1717
+ "knowledge": "2025-04",
1718
+ "release_date": "2026-04-07",
1719
+ "last_updated": "2026-04-07",
1421
1720
  "modalities": {
1422
1721
  "input": [
1423
1722
  "text"
@@ -1428,54 +1727,21 @@
1428
1727
  },
1429
1728
  "open_weights": true,
1430
1729
  "limit": {
1431
- "context": 1048576,
1730
+ "context": 202752,
1432
1731
  "output": 16384
1433
1732
  },
1434
1733
  "cost": {
1435
- "input": 1.3,
1436
- "output": 2.6,
1437
- "cache_read": 0.1
1438
- }
1439
- },
1440
- "stepfun-ai/Step-3.7-Flash": {
1441
- "id": "stepfun-ai/Step-3.7-Flash",
1442
- "name": "Step 3.7 Flash",
1443
- "description": "Newer StepFun flash model for faster agents, coding, and multimodal prompts",
1444
- "attachment": true,
1445
- "reasoning": true,
1446
- "reasoning_options": [],
1447
- "tool_call": true,
1448
- "temperature": true,
1449
- "knowledge": "2026-03-01",
1450
- "release_date": "2026-05-29",
1451
- "last_updated": "2026-05-29",
1452
- "modalities": {
1453
- "input": [
1454
- "text",
1455
- "image",
1456
- "video"
1457
- ],
1458
- "output": [
1459
- "text"
1460
- ]
1461
- },
1462
- "open_weights": true,
1463
- "limit": {
1464
- "context": 262144,
1465
- "output": 256000
1466
- },
1467
- "cost": {
1468
- "input": 0.2,
1469
- "output": 1.15,
1470
- "cache_read": 0.04
1734
+ "input": 1.05,
1735
+ "output": 3.5,
1736
+ "cache_read": 0.205
1471
1737
  }
1472
1738
  },
1473
- "moonshotai/Kimi-K2.6": {
1474
- "id": "moonshotai/Kimi-K2.6",
1475
- "name": "Kimi K2.6",
1476
- "description": "Kimi multimodal agent model for visual understanding, coding, and planning",
1477
- "family": "kimi-k2",
1478
- "attachment": true,
1739
+ "zai-org/GLM-4.6": {
1740
+ "id": "zai-org/GLM-4.6",
1741
+ "name": "GLM-4.6",
1742
+ "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
1743
+ "family": "glm",
1744
+ "attachment": false,
1479
1745
  "reasoning": true,
1480
1746
  "reasoning_options": [
1481
1747
  {
@@ -1488,14 +1754,12 @@
1488
1754
  },
1489
1755
  "structured_output": true,
1490
1756
  "temperature": true,
1491
- "knowledge": "2024-04",
1492
- "release_date": "2026-04-21",
1493
- "last_updated": "2026-04-21",
1757
+ "knowledge": "2025-04",
1758
+ "release_date": "2025-09-30",
1759
+ "last_updated": "2025-09-30",
1494
1760
  "modalities": {
1495
1761
  "input": [
1496
- "text",
1497
- "image",
1498
- "video"
1762
+ "text"
1499
1763
  ],
1500
1764
  "output": [
1501
1765
  "text"
@@ -1503,21 +1767,21 @@
1503
1767
  },
1504
1768
  "open_weights": true,
1505
1769
  "limit": {
1506
- "context": 262144,
1507
- "output": 16384
1770
+ "context": 202752,
1771
+ "output": 131072
1508
1772
  },
1509
1773
  "cost": {
1510
- "input": 0.75,
1511
- "output": 3.5,
1512
- "cache_read": 0.15
1774
+ "input": 0.5,
1775
+ "output": 2,
1776
+ "cache_read": 0.1
1513
1777
  }
1514
1778
  },
1515
- "moonshotai/Kimi-K2.5": {
1516
- "id": "moonshotai/Kimi-K2.5",
1517
- "name": "Kimi K2.5",
1518
- "description": "Kimi multimodal agent model for visual understanding, coding, and planning",
1519
- "family": "kimi-k2",
1520
- "attachment": true,
1779
+ "zai-org/GLM-5": {
1780
+ "id": "zai-org/GLM-5",
1781
+ "name": "GLM-5",
1782
+ "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
1783
+ "family": "glm",
1784
+ "attachment": false,
1521
1785
  "reasoning": true,
1522
1786
  "reasoning_options": [
1523
1787
  {
@@ -1530,14 +1794,12 @@
1530
1794
  },
1531
1795
  "structured_output": true,
1532
1796
  "temperature": true,
1533
- "knowledge": "2025-01",
1534
- "release_date": "2026-01-27",
1535
- "last_updated": "2026-01-27",
1797
+ "knowledge": "2025-12",
1798
+ "release_date": "2026-02-12",
1799
+ "last_updated": "2026-02-12",
1536
1800
  "modalities": {
1537
1801
  "input": [
1538
- "text",
1539
- "image",
1540
- "video"
1802
+ "text"
1541
1803
  ],
1542
1804
  "output": [
1543
1805
  "text"
@@ -1545,21 +1807,21 @@
1545
1807
  },
1546
1808
  "open_weights": true,
1547
1809
  "limit": {
1548
- "context": 262144,
1549
- "output": 32768
1810
+ "context": 202752,
1811
+ "output": 16384
1550
1812
  },
1551
1813
  "cost": {
1552
- "input": 0.45,
1553
- "output": 2.25,
1554
- "cache_read": 0.07
1814
+ "input": 0.6,
1815
+ "output": 2.08,
1816
+ "cache_read": 0.12
1555
1817
  }
1556
1818
  },
1557
- "moonshotai/Kimi-K2.7-Code": {
1558
- "id": "moonshotai/Kimi-K2.7-Code",
1559
- "name": "Kimi K2.7 Code",
1560
- "description": "Coding-focused Kimi model, stronger on long-horizon repo work with less overthinking",
1561
- "family": "kimi-k2",
1562
- "attachment": true,
1819
+ "zai-org/GLM-4.7": {
1820
+ "id": "zai-org/GLM-4.7",
1821
+ "name": "GLM-4.7",
1822
+ "description": "Flagship GLM model for hybrid reasoning, coding, and agentic engineering",
1823
+ "family": "glm",
1824
+ "attachment": false,
1563
1825
  "reasoning": true,
1564
1826
  "reasoning_options": [
1565
1827
  {
@@ -1571,15 +1833,13 @@
1571
1833
  "field": "reasoning_content"
1572
1834
  },
1573
1835
  "structured_output": true,
1574
- "temperature": false,
1575
- "knowledge": "2025-01",
1576
- "release_date": "2026-06-12",
1577
- "last_updated": "2026-06-12",
1836
+ "temperature": true,
1837
+ "knowledge": "2025-04",
1838
+ "release_date": "2025-12-22",
1839
+ "last_updated": "2025-12-22",
1578
1840
  "modalities": {
1579
1841
  "input": [
1580
- "text",
1581
- "image",
1582
- "video"
1842
+ "text"
1583
1843
  ],
1584
1844
  "output": [
1585
1845
  "text"
@@ -1587,41 +1847,47 @@
1587
1847
  },
1588
1848
  "open_weights": true,
1589
1849
  "limit": {
1590
- "context": 262144,
1591
- "output": 262144
1850
+ "context": 202752,
1851
+ "output": 16384
1592
1852
  },
1593
1853
  "cost": {
1594
- "input": 0.74,
1595
- "output": 3.5,
1596
- "cache_read": 0.15
1854
+ "input": 0.4,
1855
+ "output": 1.75,
1856
+ "cache_read": 0.08
1597
1857
  }
1598
1858
  },
1599
- "moonshotai/Kimi-K3": {
1600
- "id": "moonshotai/Kimi-K3",
1601
- "name": "Kimi K3",
1602
- "description": "Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work",
1603
- "family": "kimi-k3",
1604
- "attachment": true,
1859
+ "zai-org/GLM-5.2": {
1860
+ "id": "zai-org/GLM-5.2",
1861
+ "name": "GLM-5.2",
1862
+ "description": "Open flagship GLM for long-horizon coding agents and million-token context work",
1863
+ "family": "glm",
1864
+ "attachment": false,
1605
1865
  "reasoning": true,
1606
1866
  "reasoning_options": [
1867
+ {
1868
+ "type": "toggle"
1869
+ },
1607
1870
  {
1608
1871
  "type": "effort",
1609
1872
  "values": [
1610
1873
  "low",
1874
+ "medium",
1611
1875
  "high",
1612
- "max"
1876
+ "xhigh"
1613
1877
  ]
1614
1878
  }
1615
1879
  ],
1616
1880
  "tool_call": true,
1881
+ "interleaved": {
1882
+ "field": "reasoning_content"
1883
+ },
1617
1884
  "structured_output": true,
1618
- "temperature": false,
1619
- "release_date": "2026-07-16",
1620
- "last_updated": "2026-07-16",
1885
+ "temperature": true,
1886
+ "release_date": "2026-06-13",
1887
+ "last_updated": "2026-06-13",
1621
1888
  "modalities": {
1622
1889
  "input": [
1623
- "text",
1624
- "image"
1890
+ "text"
1625
1891
  ],
1626
1892
  "output": [
1627
1893
  "text"
@@ -1630,36 +1896,31 @@
1630
1896
  "open_weights": true,
1631
1897
  "limit": {
1632
1898
  "context": 1048576,
1633
- "output": 131072
1899
+ "output": 32768
1634
1900
  },
1635
1901
  "cost": {
1636
- "input": 2.7,
1637
- "output": 13.5,
1638
- "cache_read": 0.27
1902
+ "input": 0.75,
1903
+ "output": 2.4,
1904
+ "cache_read": 0.14
1639
1905
  }
1640
1906
  },
1641
- "openai/gpt-oss-20b": {
1642
- "id": "openai/gpt-oss-20b",
1643
- "name": "GPT OSS 20B",
1644
- "description": "Open-weight GPT model for self-hosted reasoning and instruction-following workloads",
1645
- "family": "gpt-oss",
1907
+ "zai-org/GLM-4.7-Flash": {
1908
+ "id": "zai-org/GLM-4.7-Flash",
1909
+ "name": "GLM-4.7-Flash",
1910
+ "description": "Efficient GLM model for fast reasoning, coding, and agent workflows",
1911
+ "family": "glm-flash",
1646
1912
  "attachment": false,
1647
1913
  "reasoning": true,
1648
- "reasoning_options": [
1649
- {
1650
- "type": "effort",
1651
- "values": [
1652
- "low",
1653
- "medium",
1654
- "high"
1655
- ]
1656
- }
1657
- ],
1914
+ "reasoning_options": [],
1658
1915
  "tool_call": true,
1916
+ "interleaved": {
1917
+ "field": "reasoning_content"
1918
+ },
1659
1919
  "structured_output": true,
1660
1920
  "temperature": true,
1661
- "release_date": "2025-08-05",
1662
- "last_updated": "2025-08-05",
1921
+ "knowledge": "2025-04",
1922
+ "release_date": "2026-01-19",
1923
+ "last_updated": "2026-01-19",
1663
1924
  "modalities": {
1664
1925
  "input": [
1665
1926
  "text"
@@ -1670,39 +1931,32 @@
1670
1931
  },
1671
1932
  "open_weights": true,
1672
1933
  "limit": {
1673
- "context": 131072,
1934
+ "context": 202752,
1674
1935
  "output": 16384
1675
1936
  },
1676
1937
  "cost": {
1677
- "input": 0.03,
1678
- "output": 0.14
1938
+ "input": 0.06,
1939
+ "output": 0.4,
1940
+ "cache_read": 0.01
1679
1941
  }
1680
1942
  },
1681
- "openai/gpt-oss-120b": {
1682
- "id": "openai/gpt-oss-120b",
1683
- "name": "GPT OSS 120B",
1684
- "description": "Open-weight GPT model for self-hosted reasoning and instruction-following workloads",
1685
- "family": "gpt-oss",
1686
- "attachment": false,
1943
+ "thinkingmachines/Inkling": {
1944
+ "id": "thinkingmachines/Inkling",
1945
+ "name": "Inkling",
1946
+ "description": "Multimodal MoE reasoning model (975B total, 41B active) for text, image, and audio",
1947
+ "family": "ling",
1948
+ "attachment": true,
1687
1949
  "reasoning": true,
1688
- "reasoning_options": [
1689
- {
1690
- "type": "effort",
1691
- "values": [
1692
- "low",
1693
- "medium",
1694
- "high"
1695
- ]
1696
- }
1697
- ],
1950
+ "reasoning_options": [],
1698
1951
  "tool_call": true,
1699
- "structured_output": true,
1700
1952
  "temperature": true,
1701
- "release_date": "2025-08-05",
1702
- "last_updated": "2025-08-05",
1953
+ "release_date": "2026-07-15",
1954
+ "last_updated": "2026-07-15",
1703
1955
  "modalities": {
1704
1956
  "input": [
1705
- "text"
1957
+ "text",
1958
+ "image",
1959
+ "audio"
1706
1960
  ],
1707
1961
  "output": [
1708
1962
  "text"
@@ -1710,29 +1964,33 @@
1710
1964
  },
1711
1965
  "open_weights": true,
1712
1966
  "limit": {
1713
- "context": 131072,
1714
- "output": 16384
1967
+ "context": 524288,
1968
+ "output": 1048576
1715
1969
  },
1716
1970
  "cost": {
1717
- "input": 0.037,
1718
- "output": 0.17
1971
+ "input": 0.95,
1972
+ "output": 4.05,
1973
+ "cache_read": 0.16
1719
1974
  }
1720
1975
  },
1721
- "meta-llama/Llama-4-Scout-17B-16E-Instruct": {
1722
- "id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
1723
- "name": "Llama 4 Scout 17B",
1724
- "description": "Open multimodal Llama model for long-context analysis and efficient agents",
1725
- "family": "llama",
1976
+ "thinkingmachines/Inkling-Small": {
1977
+ "id": "thinkingmachines/Inkling-Small",
1978
+ "name": "Inkling Small",
1979
+ "description": "Multimodal MoE reasoning model (276B total, 12B active) for text, image, and audio",
1980
+ "family": "ling",
1726
1981
  "attachment": true,
1727
- "reasoning": false,
1982
+ "reasoning": true,
1983
+ "reasoning_options": [],
1728
1984
  "tool_call": true,
1729
1985
  "structured_output": true,
1730
- "release_date": "2025-04-05",
1731
- "last_updated": "2025-04-05",
1986
+ "temperature": true,
1987
+ "release_date": "2026-07-30",
1988
+ "last_updated": "2026-07-30",
1732
1989
  "modalities": {
1733
1990
  "input": [
1734
1991
  "text",
1735
- "image"
1992
+ "image",
1993
+ "audio"
1736
1994
  ],
1737
1995
  "output": [
1738
1996
  "text"
@@ -1740,12 +1998,13 @@
1740
1998
  },
1741
1999
  "open_weights": true,
1742
2000
  "limit": {
1743
- "context": 327680,
1744
- "output": 16384
2001
+ "context": 524288,
2002
+ "output": 1048576
1745
2003
  },
1746
2004
  "cost": {
1747
- "input": 0.1,
1748
- "output": 0.3
2005
+ "input": 0.45,
2006
+ "output": 1.2,
2007
+ "cache_read": 0.1
1749
2008
  }
1750
2009
  },
1751
2010
  "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8": {
@@ -1807,74 +2066,21 @@
1807
2066
  "output": 0.32
1808
2067
  }
1809
2068
  },
1810
- "XiaomiMiMo/MiMo-V2.5-Pro": {
1811
- "id": "XiaomiMiMo/MiMo-V2.5-Pro",
1812
- "name": "MiMo-V2.5-Pro",
1813
- "description": "Stronger MiMo Pro tier for multimodal reasoning and coding-agent execution",
1814
- "family": "mimo",
1815
- "attachment": true,
1816
- "reasoning": true,
1817
- "reasoning_options": [
1818
- {
1819
- "type": "toggle"
1820
- }
1821
- ],
1822
- "tool_call": true,
1823
- "interleaved": {
1824
- "field": "reasoning_content"
1825
- },
1826
- "structured_output": true,
1827
- "temperature": true,
1828
- "knowledge": "2024-12",
1829
- "release_date": "2026-04-22",
1830
- "last_updated": "2026-04-22",
1831
- "modalities": {
1832
- "input": [
1833
- "text",
1834
- "audio"
1835
- ],
1836
- "output": [
1837
- "text"
1838
- ]
1839
- },
1840
- "open_weights": true,
1841
- "limit": {
1842
- "context": 1048576,
1843
- "output": 16384
1844
- },
1845
- "cost": {
1846
- "input": 1,
1847
- "output": 3,
1848
- "cache_read": 0.2
1849
- }
1850
- },
1851
- "XiaomiMiMo/MiMo-V2.5": {
1852
- "id": "XiaomiMiMo/MiMo-V2.5",
1853
- "name": "MiMo-V2.5",
1854
- "description": "Open MiMo model for multimodal coding agents and long-context automation",
1855
- "family": "mimo",
2069
+ "meta-llama/Llama-4-Scout-17B-16E-Instruct": {
2070
+ "id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
2071
+ "name": "Llama 4 Scout 17B",
2072
+ "description": "Open multimodal Llama model for long-context analysis and efficient agents",
2073
+ "family": "llama",
1856
2074
  "attachment": true,
1857
- "reasoning": true,
1858
- "reasoning_options": [
1859
- {
1860
- "type": "toggle"
1861
- }
1862
- ],
2075
+ "reasoning": false,
1863
2076
  "tool_call": true,
1864
- "interleaved": {
1865
- "field": "reasoning_content"
1866
- },
1867
2077
  "structured_output": true,
1868
- "temperature": true,
1869
- "knowledge": "2024-12",
1870
- "release_date": "2026-04-22",
1871
- "last_updated": "2026-04-22",
2078
+ "release_date": "2025-04-05",
2079
+ "last_updated": "2025-04-05",
1872
2080
  "modalities": {
1873
2081
  "input": [
1874
2082
  "text",
1875
- "image",
1876
- "audio",
1877
- "video"
2083
+ "image"
1878
2084
  ],
1879
2085
  "output": [
1880
2086
  "text"
@@ -1882,13 +2088,12 @@
1882
2088
  },
1883
2089
  "open_weights": true,
1884
2090
  "limit": {
1885
- "context": 262144,
2091
+ "context": 327680,
1886
2092
  "output": 16384
1887
2093
  },
1888
2094
  "cost": {
1889
- "input": 0.4,
1890
- "output": 2,
1891
- "cache_read": 0.08
2095
+ "input": 0.1,
2096
+ "output": 0.3
1892
2097
  }
1893
2098
  }
1894
2099
  }