llm.rb 15.0.3 → 15.2.0

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 (104) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +433 -3
  3. data/README.md +188 -71
  4. data/bin/llm.rb +50 -7
  5. data/data/alibaba.json +912 -823
  6. data/data/anthropic.json +234 -187
  7. data/data/bedrock.json +3702 -2058
  8. data/data/deepinfra.json +1288 -951
  9. data/data/deepseek.json +87 -53
  10. data/data/google.json +670 -670
  11. data/data/mistral.json +501 -460
  12. data/data/moonshot.json +43 -248
  13. data/data/openai.json +1008 -914
  14. data/data/openrouter.json +14417 -0
  15. data/data/xai.json +213 -201
  16. data/data/zai.json +242 -149
  17. data/docs/deepdive/advanced/compaction.md +5 -5
  18. data/docs/deepdive/advanced/context.md +8 -6
  19. data/docs/deepdive/advanced/guard.md +2 -2
  20. data/docs/deepdive/features/builtin_tools.md +93 -22
  21. data/docs/deepdive/features/{repl.md → console.md} +28 -28
  22. data/docs/deepdive/features/database.md +3 -3
  23. data/docs/deepdive/fundamentals/agents.md +13 -12
  24. data/docs/deepdive/fundamentals/providers.md +91 -6
  25. data/docs/deepdive/fundamentals/skills.md +14 -6
  26. data/docs/deepdive/fundamentals/stream.md +4 -4
  27. data/docs/deepdive/fundamentals/tools.md +63 -31
  28. data/docs/deepdive/reference/cost.md +2 -2
  29. data/docs/deepdive/reference/model_registry.md +2 -2
  30. data/docs/deepdive/reference/tracer.md +15 -13
  31. data/docs/deepdive.md +2 -2
  32. data/lib/llm/active_record/acts_as_agent.rb +9 -5
  33. data/lib/llm/agent.rb +40 -15
  34. data/lib/llm/{repl → console}/bar.rb +3 -3
  35. data/lib/llm/{repl → console}/buffer.rb +24 -9
  36. data/lib/llm/{repl → console}/color.rb +2 -2
  37. data/lib/llm/{repl → console}/command.rb +12 -12
  38. data/lib/llm/{repl → console}/commands/exit.rb +4 -4
  39. data/lib/llm/{repl → console}/commands/help.rb +1 -1
  40. data/lib/llm/{repl/commands/compact.rb → console/commands/keep.rb} +11 -9
  41. data/lib/llm/{repl → console}/commands/model.rb +2 -2
  42. data/lib/llm/{repl → console}/input/cache.rb +2 -2
  43. data/lib/llm/{repl → console}/input/char.rb +2 -2
  44. data/lib/llm/{repl → console}/input/row.rb +1 -1
  45. data/lib/llm/{repl → console}/input.rb +18 -10
  46. data/lib/llm/console/markdown/parser.rb +78 -0
  47. data/lib/llm/{repl → console}/markdown/table.rb +8 -5
  48. data/lib/llm/{repl → console}/markdown.rb +13 -30
  49. data/lib/llm/console/node.rb +69 -0
  50. data/lib/llm/{repl → console}/status.rb +11 -11
  51. data/lib/llm/{repl → console}/stream.rb +36 -9
  52. data/lib/llm/{repl → console}/walker.rb +1 -1
  53. data/lib/llm/{repl → console}/window.rb +17 -17
  54. data/lib/llm/{repl.rb → console.rb} +39 -19
  55. data/lib/llm/context/deserializer.rb +2 -1
  56. data/lib/llm/context.rb +29 -13
  57. data/lib/llm/cost.rb +13 -0
  58. data/lib/llm/function/async/reactor.rb +20 -1
  59. data/lib/llm/function/fork/task.rb +14 -10
  60. data/lib/llm/function.rb +1 -1
  61. data/lib/llm/json_adapter.rb +40 -28
  62. data/lib/llm/message.rb +7 -0
  63. data/lib/llm/provider.rb +31 -10
  64. data/lib/llm/providers/alibaba.rb +1 -1
  65. data/lib/llm/providers/anthropic.rb +1 -1
  66. data/lib/llm/providers/bedrock/models.rb +2 -2
  67. data/lib/llm/providers/bedrock.rb +1 -1
  68. data/lib/llm/providers/deepseek.rb +1 -1
  69. data/lib/llm/providers/google.rb +1 -1
  70. data/lib/llm/providers/ollama.rb +1 -1
  71. data/lib/llm/providers/openai/responses.rb +2 -1
  72. data/lib/llm/providers/openai.rb +2 -1
  73. data/lib/llm/providers/openrouter.rb +87 -0
  74. data/lib/llm/schema/leaf.rb +34 -2
  75. data/lib/llm/schema.rb +4 -2
  76. data/lib/llm/sequel/agent.rb +9 -5
  77. data/lib/llm/skill.rb +7 -1
  78. data/lib/llm/stream.rb +8 -3
  79. data/lib/llm/tool/param.rb +5 -1
  80. data/lib/llm/tool.rb +5 -0
  81. data/lib/llm/tools/bundle.rb +53 -0
  82. data/lib/llm/tools/edit-file.rb +7 -2
  83. data/lib/llm/tools/exec.rb +78 -0
  84. data/lib/llm/tools/git.rb +27 -26
  85. data/lib/llm/tools/mkdir.rb +12 -19
  86. data/lib/llm/tools/read_file.rb +69 -9
  87. data/lib/llm/tools/rg.rb +20 -24
  88. data/lib/llm/tools/ruby.rb +17 -25
  89. data/lib/llm/tools/utils.rb +75 -2
  90. data/lib/llm/tools/write_file.rb +4 -1
  91. data/lib/llm/tracer/logger.rb +2 -2
  92. data/lib/llm/tracer/pretty_logger.rb +4 -4
  93. data/lib/llm/tracer/telemetry.rb +2 -2
  94. data/lib/llm/tracer.rb +33 -0
  95. data/lib/llm/transport/curb.rb +5 -3
  96. data/lib/llm/transport/http.rb +5 -2
  97. data/lib/llm/transport/persistent_http.rb +6 -4
  98. data/lib/llm/transport/utils.rb +8 -6
  99. data/lib/llm/version.rb +1 -1
  100. data/lib/llm.rb +18 -12
  101. data/llm.gemspec +8 -8
  102. metadata +80 -37
  103. data/lib/llm/repl/node.rb +0 -44
  104. data/lib/llm/tools/shell.rb +0 -55
data/data/alibaba.json CHANGED
@@ -8,10 +8,49 @@
8
8
  "name": "Alibaba",
9
9
  "doc": "https://www.alibabacloud.com/help/en/model-studio/models",
10
10
  "models": {
11
- "qwen2-5-vl-7b-instruct": {
12
- "id": "qwen2-5-vl-7b-instruct",
13
- "name": "Qwen2.5-VL 7B Instruct",
14
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
11
+ "qwen3.7-max": {
12
+ "id": "qwen3.7-max",
13
+ "name": "Qwen3.7 Max",
14
+ "description": "Qwen frontier model tuned for agent frameworks, coding assistants, and long tasks",
15
+ "family": "qwen",
16
+ "attachment": false,
17
+ "reasoning": true,
18
+ "reasoning_options": [
19
+ {
20
+ "type": "toggle"
21
+ },
22
+ {
23
+ "type": "budget_tokens"
24
+ }
25
+ ],
26
+ "tool_call": true,
27
+ "temperature": true,
28
+ "release_date": "2026-05-21",
29
+ "last_updated": "2026-05-21",
30
+ "modalities": {
31
+ "input": [
32
+ "text"
33
+ ],
34
+ "output": [
35
+ "text"
36
+ ]
37
+ },
38
+ "open_weights": false,
39
+ "limit": {
40
+ "context": 1000000,
41
+ "output": 65536
42
+ },
43
+ "cost": {
44
+ "input": 2.5,
45
+ "output": 7.5,
46
+ "cache_read": 0.5,
47
+ "cache_write": 3.125
48
+ }
49
+ },
50
+ "qwen2-5-72b-instruct": {
51
+ "id": "qwen2-5-72b-instruct",
52
+ "name": "Qwen2.5 72B Instruct",
53
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
15
54
  "family": "qwen",
16
55
  "attachment": false,
17
56
  "reasoning": false,
@@ -22,8 +61,7 @@
22
61
  "last_updated": "2024-09",
23
62
  "modalities": {
24
63
  "input": [
25
- "text",
26
- "image"
64
+ "text"
27
65
  ],
28
66
  "output": [
29
67
  "text"
@@ -35,15 +73,15 @@
35
73
  "output": 8192
36
74
  },
37
75
  "cost": {
38
- "input": 0.35,
39
- "output": 1.05
76
+ "input": 1.4,
77
+ "output": 5.6
40
78
  }
41
79
  },
42
- "qwen3-vl-plus": {
43
- "id": "qwen3-vl-plus",
44
- "name": "Qwen3-VL Plus",
45
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
46
- "family": "qwen",
80
+ "deepseek-v4-flash-0731": {
81
+ "id": "deepseek-v4-flash-0731",
82
+ "name": "DeepSeek V4 Flash 0731",
83
+ "description": "Official DeepSeek V4 Flash release with enhanced agentic capabilities and integrated DSpark speculative decoding",
84
+ "family": "deepseek-flash",
47
85
  "attachment": false,
48
86
  "reasoning": true,
49
87
  "reasoning_options": [
@@ -51,46 +89,53 @@
51
89
  "type": "toggle"
52
90
  },
53
91
  {
54
- "type": "budget_tokens"
92
+ "type": "effort",
93
+ "values": [
94
+ "high",
95
+ "max"
96
+ ]
55
97
  }
56
98
  ],
57
99
  "tool_call": true,
100
+ "interleaved": {
101
+ "field": "reasoning_content"
102
+ },
103
+ "structured_output": true,
58
104
  "temperature": true,
59
- "knowledge": "2025-04",
60
- "release_date": "2025-09-23",
61
- "last_updated": "2025-09-23",
105
+ "knowledge": "2025-05",
106
+ "release_date": "2026-07-31",
107
+ "last_updated": "2026-07-31",
62
108
  "modalities": {
63
109
  "input": [
64
- "text",
65
- "image"
110
+ "text"
66
111
  ],
67
112
  "output": [
68
113
  "text"
69
114
  ]
70
115
  },
71
- "open_weights": false,
116
+ "open_weights": true,
72
117
  "limit": {
73
- "context": 262144,
74
- "output": 32768
118
+ "context": 1000000,
119
+ "output": 384000
75
120
  },
76
121
  "cost": {
77
122
  "input": 0.2,
78
- "output": 1.6,
79
- "reasoning": 4.8
123
+ "output": 0.4,
124
+ "cache_read": 0.04
80
125
  }
81
126
  },
82
- "qwen3-coder-30b-a3b-instruct": {
83
- "id": "qwen3-coder-30b-a3b-instruct",
84
- "name": "Qwen3-Coder 30B-A3B Instruct",
85
- "description": "Smaller Qwen coder for efficient local agents and repo-level fixes",
127
+ "qwen3-coder-plus": {
128
+ "id": "qwen3-coder-plus",
129
+ "name": "Qwen3 Coder Plus",
130
+ "description": "Hosted Qwen coder for software agents, repo edits, and long-context code",
86
131
  "family": "qwen",
87
132
  "attachment": false,
88
133
  "reasoning": false,
89
134
  "tool_call": true,
90
135
  "temperature": true,
91
136
  "knowledge": "2025-04",
92
- "release_date": "2025-04",
93
- "last_updated": "2025-04",
137
+ "release_date": "2025-07-23",
138
+ "last_updated": "2025-07-23",
94
139
  "modalities": {
95
140
  "input": [
96
141
  "text"
@@ -101,33 +146,31 @@
101
146
  },
102
147
  "open_weights": true,
103
148
  "limit": {
104
- "context": 262144,
149
+ "context": 1048576,
105
150
  "output": 65536
106
151
  },
107
152
  "cost": {
108
- "input": 0.45,
109
- "output": 2.25
153
+ "input": 1,
154
+ "output": 5
110
155
  }
111
156
  },
112
- "qwen3.7-max": {
113
- "id": "qwen3.7-max",
114
- "name": "Qwen3.7 Max",
115
- "description": "Qwen frontier model tuned for agent frameworks, coding assistants, and long tasks",
157
+ "qwen3-next-80b-a3b-thinking": {
158
+ "id": "qwen3-next-80b-a3b-thinking",
159
+ "name": "Qwen3-Next 80B-A3B (Thinking)",
160
+ "description": "Efficient Qwen thinking model for local reasoning, math, and coding agents",
116
161
  "family": "qwen",
117
162
  "attachment": false,
118
163
  "reasoning": true,
119
164
  "reasoning_options": [
120
- {
121
- "type": "toggle"
122
- },
123
165
  {
124
166
  "type": "budget_tokens"
125
167
  }
126
168
  ],
127
169
  "tool_call": true,
128
170
  "temperature": true,
129
- "release_date": "2026-05-21",
130
- "last_updated": "2026-05-21",
171
+ "knowledge": "2025-04",
172
+ "release_date": "2025-09",
173
+ "last_updated": "2025-09",
131
174
  "modalities": {
132
175
  "input": [
133
176
  "text"
@@ -136,91 +179,79 @@
136
179
  "text"
137
180
  ]
138
181
  },
139
- "open_weights": false,
182
+ "open_weights": true,
140
183
  "limit": {
141
- "context": 1000000,
142
- "output": 65536
184
+ "context": 131072,
185
+ "output": 32768
143
186
  },
144
187
  "cost": {
145
- "input": 2.5,
146
- "output": 7.5,
147
- "cache_read": 0.5,
148
- "cache_write": 3.125
188
+ "input": 0.5,
189
+ "output": 6
149
190
  }
150
191
  },
151
- "qwen-turbo": {
152
- "id": "qwen-turbo",
153
- "name": "Qwen Turbo",
154
- "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
192
+ "qwen2-5-omni-7b": {
193
+ "id": "qwen2-5-omni-7b",
194
+ "name": "Qwen2.5-Omni 7B",
195
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
155
196
  "family": "qwen",
156
197
  "attachment": false,
157
- "reasoning": true,
158
- "reasoning_options": [
159
- {
160
- "type": "toggle"
161
- },
162
- {
163
- "type": "budget_tokens"
164
- }
165
- ],
198
+ "reasoning": false,
166
199
  "tool_call": true,
167
200
  "temperature": true,
168
201
  "knowledge": "2024-04",
169
- "release_date": "2024-11-01",
170
- "last_updated": "2025-04-28",
202
+ "release_date": "2024-12",
203
+ "last_updated": "2024-12",
171
204
  "modalities": {
172
205
  "input": [
173
- "text"
206
+ "text",
207
+ "image",
208
+ "audio",
209
+ "video"
174
210
  ],
175
211
  "output": [
176
- "text"
212
+ "text",
213
+ "audio"
177
214
  ]
178
215
  },
179
- "open_weights": false,
216
+ "open_weights": true,
180
217
  "limit": {
181
- "context": 1000000,
182
- "output": 16384
218
+ "context": 32768,
219
+ "output": 2048
183
220
  },
184
221
  "cost": {
185
- "input": 0.05,
186
- "output": 0.2,
187
- "reasoning": 0.5
222
+ "input": 0.1,
223
+ "output": 0.4,
224
+ "input_audio": 6.76
188
225
  }
189
226
  },
190
- "qwen-omni-turbo": {
191
- "id": "qwen-omni-turbo",
192
- "name": "Qwen-Omni Turbo",
193
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
227
+ "qwen-mt-turbo": {
228
+ "id": "qwen-mt-turbo",
229
+ "name": "Qwen-MT Turbo",
230
+ "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
194
231
  "family": "qwen",
195
232
  "attachment": false,
196
233
  "reasoning": false,
197
- "tool_call": true,
234
+ "tool_call": false,
198
235
  "temperature": true,
199
236
  "knowledge": "2024-04",
200
- "release_date": "2025-01-19",
201
- "last_updated": "2025-03-26",
237
+ "release_date": "2025-01",
238
+ "last_updated": "2025-01",
202
239
  "modalities": {
203
240
  "input": [
204
- "text",
205
- "image",
206
- "audio",
207
- "video"
241
+ "text"
208
242
  ],
209
243
  "output": [
210
- "text",
211
- "audio"
244
+ "text"
212
245
  ]
213
246
  },
214
247
  "open_weights": false,
215
248
  "limit": {
216
- "context": 32768,
217
- "output": 2048
249
+ "context": 16384,
250
+ "output": 8192
218
251
  },
219
252
  "cost": {
220
- "input": 0.07,
221
- "output": 0.27,
222
- "input_audio": 4.44,
223
- "output_audio": 8.89
253
+ "input": 0.16,
254
+ "output": 0.49
224
255
  }
225
256
  },
226
257
  "qwen-vl-max": {
@@ -254,26 +285,18 @@
254
285
  "output": 3.2
255
286
  }
256
287
  },
257
- "qwen3-32b": {
258
- "id": "qwen3-32b",
259
- "name": "Qwen3 32B",
260
- "description": "Dense open Qwen model for self-hosted chat, reasoning, and coding",
288
+ "qwen3-next-80b-a3b-instruct": {
289
+ "id": "qwen3-next-80b-a3b-instruct",
290
+ "name": "Qwen3-Next 80B-A3B Instruct",
291
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
261
292
  "family": "qwen",
262
293
  "attachment": false,
263
- "reasoning": true,
264
- "reasoning_options": [
265
- {
266
- "type": "toggle"
267
- },
268
- {
269
- "type": "budget_tokens"
270
- }
271
- ],
294
+ "reasoning": false,
272
295
  "tool_call": true,
273
296
  "temperature": true,
274
297
  "knowledge": "2025-04",
275
- "release_date": "2025-04",
276
- "last_updated": "2025-04",
298
+ "release_date": "2025-09",
299
+ "last_updated": "2025-09",
277
300
  "modalities": {
278
301
  "input": [
279
302
  "text"
@@ -285,55 +308,47 @@
285
308
  "open_weights": true,
286
309
  "limit": {
287
310
  "context": 131072,
288
- "output": 16384
311
+ "output": 32768
289
312
  },
290
313
  "cost": {
291
- "input": 0.7,
292
- "output": 2.8,
293
- "reasoning": 8.4
314
+ "input": 0.5,
315
+ "output": 2
294
316
  }
295
317
  },
296
- "qwen3-vl-30b-a3b": {
297
- "id": "qwen3-vl-30b-a3b",
298
- "name": "Qwen3-VL 30B-A3B",
299
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
318
+ "qwen3-coder-flash": {
319
+ "id": "qwen3-coder-flash",
320
+ "name": "Qwen3 Coder Flash",
321
+ "description": "Qwen coding model for software agents, repository edits, and code reasoning",
300
322
  "family": "qwen",
301
323
  "attachment": false,
302
- "reasoning": true,
303
- "reasoning_options": [
304
- {
305
- "type": "budget_tokens"
306
- }
307
- ],
324
+ "reasoning": false,
308
325
  "tool_call": true,
309
326
  "temperature": true,
310
327
  "knowledge": "2025-04",
311
- "release_date": "2025-04",
312
- "last_updated": "2025-04",
328
+ "release_date": "2025-07-28",
329
+ "last_updated": "2025-07-28",
313
330
  "modalities": {
314
331
  "input": [
315
- "text",
316
- "image"
332
+ "text"
317
333
  ],
318
334
  "output": [
319
335
  "text"
320
336
  ]
321
337
  },
322
- "open_weights": true,
338
+ "open_weights": false,
323
339
  "limit": {
324
- "context": 131072,
325
- "output": 32768
340
+ "context": 1000000,
341
+ "output": 65536
326
342
  },
327
343
  "cost": {
328
- "input": 0.2,
329
- "output": 0.8,
330
- "reasoning": 2.4
344
+ "input": 0.3,
345
+ "output": 1.5
331
346
  }
332
347
  },
333
- "qwen3-235b-a22b": {
334
- "id": "qwen3-235b-a22b",
335
- "name": "Qwen3 235B-A22B",
336
- "description": "Large open Qwen MoE for multilingual reasoning, coding, and tool use",
348
+ "qwen3-14b": {
349
+ "id": "qwen3-14b",
350
+ "name": "Qwen3 14B",
351
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
337
352
  "family": "qwen",
338
353
  "attachment": false,
339
354
  "reasoning": true,
@@ -361,12 +376,12 @@
361
376
  "open_weights": true,
362
377
  "limit": {
363
378
  "context": 131072,
364
- "output": 16384
379
+ "output": 8192
365
380
  },
366
381
  "cost": {
367
- "input": 0.7,
368
- "output": 2.8,
369
- "reasoning": 8.4
382
+ "input": 0.35,
383
+ "output": 1.4,
384
+ "reasoning": 4.2
370
385
  }
371
386
  },
372
387
  "qwen-max": {
@@ -399,9 +414,135 @@
399
414
  "output": 6.4
400
415
  }
401
416
  },
402
- "qwen3.5-35b-a3b": {
403
- "id": "qwen3.5-35b-a3b",
404
- "name": "Qwen3.5 35B-A3B",
417
+ "qwen3.6-plus": {
418
+ "id": "qwen3.6-plus",
419
+ "name": "Qwen3.6 Plus",
420
+ "description": "Earlier Qwen multimodal workhorse for million-token agent and document tasks",
421
+ "family": "qwen",
422
+ "attachment": false,
423
+ "reasoning": true,
424
+ "reasoning_options": [
425
+ {
426
+ "type": "toggle"
427
+ },
428
+ {
429
+ "type": "budget_tokens"
430
+ }
431
+ ],
432
+ "tool_call": true,
433
+ "temperature": true,
434
+ "knowledge": "2025-04",
435
+ "release_date": "2026-04-02",
436
+ "last_updated": "2026-04-02",
437
+ "modalities": {
438
+ "input": [
439
+ "text",
440
+ "image",
441
+ "video"
442
+ ],
443
+ "output": [
444
+ "text"
445
+ ]
446
+ },
447
+ "open_weights": false,
448
+ "limit": {
449
+ "context": 1000000,
450
+ "output": 65536
451
+ },
452
+ "cost": {
453
+ "input": 0.5,
454
+ "output": 3,
455
+ "cache_read": 0.05,
456
+ "cache_write": 0.625,
457
+ "tiers": [
458
+ {
459
+ "input": 2,
460
+ "output": 6,
461
+ "cache_read": 0.2,
462
+ "cache_write": 2.5,
463
+ "tier": {
464
+ "type": "context",
465
+ "size": 256000
466
+ }
467
+ }
468
+ ],
469
+ "context_over_200k": {
470
+ "input": 2,
471
+ "output": 6,
472
+ "cache_read": 0.2,
473
+ "cache_write": 2.5
474
+ }
475
+ }
476
+ },
477
+ "qwen-vl-plus": {
478
+ "id": "qwen-vl-plus",
479
+ "name": "Qwen-VL Plus",
480
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
481
+ "family": "qwen",
482
+ "attachment": false,
483
+ "reasoning": false,
484
+ "tool_call": true,
485
+ "temperature": true,
486
+ "knowledge": "2024-04",
487
+ "release_date": "2024-01-25",
488
+ "last_updated": "2025-08-15",
489
+ "modalities": {
490
+ "input": [
491
+ "text",
492
+ "image"
493
+ ],
494
+ "output": [
495
+ "text"
496
+ ]
497
+ },
498
+ "open_weights": false,
499
+ "limit": {
500
+ "context": 131072,
501
+ "output": 8192
502
+ },
503
+ "cost": {
504
+ "input": 0.21,
505
+ "output": 0.63
506
+ }
507
+ },
508
+ "qwen-omni-turbo-realtime": {
509
+ "id": "qwen-omni-turbo-realtime",
510
+ "name": "Qwen-Omni Turbo Realtime",
511
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
512
+ "family": "qwen",
513
+ "attachment": false,
514
+ "reasoning": false,
515
+ "tool_call": true,
516
+ "temperature": true,
517
+ "knowledge": "2024-04",
518
+ "release_date": "2025-05-08",
519
+ "last_updated": "2025-05-08",
520
+ "modalities": {
521
+ "input": [
522
+ "text",
523
+ "image",
524
+ "audio"
525
+ ],
526
+ "output": [
527
+ "text",
528
+ "audio"
529
+ ]
530
+ },
531
+ "open_weights": false,
532
+ "limit": {
533
+ "context": 32768,
534
+ "output": 2048
535
+ },
536
+ "cost": {
537
+ "input": 0.27,
538
+ "output": 1.07,
539
+ "input_audio": 4.44,
540
+ "output_audio": 8.89
541
+ }
542
+ },
543
+ "qwen3.5-27b": {
544
+ "id": "qwen3.5-27b",
545
+ "name": "Qwen3.5 27B",
405
546
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
406
547
  "family": "qwen",
407
548
  "attachment": true,
@@ -436,14 +577,14 @@
436
577
  "output": 65536
437
578
  },
438
579
  "cost": {
439
- "input": 0.25,
440
- "output": 2
580
+ "input": 0.3,
581
+ "output": 2.4
441
582
  }
442
583
  },
443
- "qwen3.5-397b-a17b": {
444
- "id": "qwen3.5-397b-a17b",
445
- "name": "Qwen3.5 397B-A17B",
446
- "description": "Large open Qwen multimodal MoE for visual agents and long technical tasks",
584
+ "qwen3.5-35b-a3b": {
585
+ "id": "qwen3.5-35b-a3b",
586
+ "name": "Qwen3.5 35B-A3B",
587
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
447
588
  "family": "qwen",
448
589
  "attachment": true,
449
590
  "reasoning": true,
@@ -458,8 +599,8 @@
458
599
  "tool_call": true,
459
600
  "structured_output": true,
460
601
  "temperature": true,
461
- "release_date": "2026-02-15",
462
- "last_updated": "2026-02-15",
602
+ "release_date": "2026-02-23",
603
+ "last_updated": "2026-02-23",
463
604
  "modalities": {
464
605
  "input": [
465
606
  "text",
@@ -477,14 +618,14 @@
477
618
  "output": 65536
478
619
  },
479
620
  "cost": {
480
- "input": 0.6,
481
- "output": 3.6
621
+ "input": 0.25,
622
+ "output": 2
482
623
  }
483
624
  },
484
- "qwen-plus": {
485
- "id": "qwen-plus",
486
- "name": "Qwen Plus",
487
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
625
+ "qwen-flash": {
626
+ "id": "qwen-flash",
627
+ "name": "Qwen Flash",
628
+ "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
488
629
  "family": "qwen",
489
630
  "attachment": false,
490
631
  "reasoning": true,
@@ -499,8 +640,8 @@
499
640
  "tool_call": true,
500
641
  "temperature": true,
501
642
  "knowledge": "2024-04",
502
- "release_date": "2024-01-25",
503
- "last_updated": "2025-09-11",
643
+ "release_date": "2025-07-28",
644
+ "last_updated": "2025-07-28",
504
645
  "modalities": {
505
646
  "input": [
506
647
  "text"
@@ -515,81 +656,96 @@
515
656
  "output": 32768
516
657
  },
517
658
  "cost": {
518
- "input": 0.4,
519
- "output": 1.2,
520
- "reasoning": 4
659
+ "input": 0.05,
660
+ "output": 0.4
521
661
  }
522
662
  },
523
- "qwen3.5-plus": {
524
- "id": "qwen3.5-plus",
525
- "name": "Qwen3.5 Plus",
526
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
527
- "family": "qwen",
663
+ "glm-5.2": {
664
+ "id": "glm-5.2",
665
+ "name": "GLM-5.2",
666
+ "description": "Open flagship GLM for long-horizon coding agents and million-token context work",
667
+ "family": "glm",
528
668
  "attachment": false,
529
669
  "reasoning": true,
530
670
  "reasoning_options": [
531
671
  {
532
- "type": "toggle"
533
- },
534
- {
535
- "type": "budget_tokens"
672
+ "type": "effort",
673
+ "values": [
674
+ "none",
675
+ "minimal",
676
+ "low",
677
+ "medium",
678
+ "high",
679
+ "xhigh",
680
+ "max"
681
+ ]
536
682
  }
537
683
  ],
538
684
  "tool_call": true,
685
+ "interleaved": {
686
+ "field": "reasoning_content"
687
+ },
688
+ "structured_output": true,
539
689
  "temperature": true,
540
- "knowledge": "2025-04",
541
- "release_date": "2026-02-16",
542
- "last_updated": "2026-02-16",
690
+ "release_date": "2026-06-13",
691
+ "last_updated": "2026-06-13",
543
692
  "modalities": {
544
693
  "input": [
545
- "text",
546
- "image",
547
- "video"
694
+ "text"
548
695
  ],
549
696
  "output": [
550
697
  "text"
551
698
  ]
552
699
  },
553
- "open_weights": false,
700
+ "open_weights": true,
554
701
  "limit": {
555
702
  "context": 1000000,
556
- "output": 65536
703
+ "output": 131072
557
704
  },
558
705
  "cost": {
559
- "input": 0.4,
560
- "output": 2.4,
561
- "reasoning": 2.4
706
+ "input": 1.4,
707
+ "output": 4.4,
708
+ "cache_read": 0.28,
709
+ "cache_write": 0
562
710
  }
563
711
  },
564
- "qwen2-5-vl-72b-instruct": {
565
- "id": "qwen2-5-vl-72b-instruct",
566
- "name": "Qwen2.5-VL 72B Instruct",
567
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
568
- "family": "qwen",
712
+ "qwen-turbo": {
713
+ "id": "qwen-turbo",
714
+ "name": "Qwen Turbo",
715
+ "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
716
+ "family": "qwen",
569
717
  "attachment": false,
570
- "reasoning": false,
718
+ "reasoning": true,
719
+ "reasoning_options": [
720
+ {
721
+ "type": "toggle"
722
+ },
723
+ {
724
+ "type": "budget_tokens"
725
+ }
726
+ ],
571
727
  "tool_call": true,
572
728
  "temperature": true,
573
729
  "knowledge": "2024-04",
574
- "release_date": "2024-09",
575
- "last_updated": "2024-09",
730
+ "release_date": "2024-11-01",
731
+ "last_updated": "2025-04-28",
576
732
  "modalities": {
577
733
  "input": [
578
- "text",
579
- "image"
734
+ "text"
580
735
  ],
581
736
  "output": [
582
737
  "text"
583
738
  ]
584
739
  },
585
- "open_weights": true,
740
+ "open_weights": false,
586
741
  "limit": {
587
- "context": 131072,
588
- "output": 8192
742
+ "context": 1000000,
743
+ "output": 16384
589
744
  },
590
745
  "cost": {
591
- "input": 2.8,
592
- "output": 8.4
746
+ "input": 0.05,
747
+ "output": 0.2,
748
+ "reasoning": 0.5
593
749
  }
594
750
  },
595
751
  "qwen3-livetranslate-flash-realtime": {
@@ -628,59 +784,59 @@
628
784
  "output_audio": 38
629
785
  }
630
786
  },
631
- "qwen3-coder-flash": {
632
- "id": "qwen3-coder-flash",
633
- "name": "Qwen3 Coder Flash",
634
- "description": "Qwen coding model for software agents, repository edits, and code reasoning",
787
+ "qwen3-vl-30b-a3b": {
788
+ "id": "qwen3-vl-30b-a3b",
789
+ "name": "Qwen3-VL 30B-A3B",
790
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
635
791
  "family": "qwen",
636
792
  "attachment": false,
637
- "reasoning": false,
793
+ "reasoning": true,
794
+ "reasoning_options": [
795
+ {
796
+ "type": "budget_tokens"
797
+ }
798
+ ],
638
799
  "tool_call": true,
639
800
  "temperature": true,
640
801
  "knowledge": "2025-04",
641
- "release_date": "2025-07-28",
642
- "last_updated": "2025-07-28",
802
+ "release_date": "2025-04",
803
+ "last_updated": "2025-04",
643
804
  "modalities": {
644
805
  "input": [
645
- "text"
806
+ "text",
807
+ "image"
646
808
  ],
647
809
  "output": [
648
810
  "text"
649
811
  ]
650
812
  },
651
- "open_weights": false,
813
+ "open_weights": true,
652
814
  "limit": {
653
- "context": 1000000,
654
- "output": 65536
815
+ "context": 131072,
816
+ "output": 32768
655
817
  },
656
818
  "cost": {
657
- "input": 0.3,
658
- "output": 1.5
819
+ "input": 0.2,
820
+ "output": 0.8,
821
+ "reasoning": 2.4
659
822
  }
660
823
  },
661
- "qwen3.6-max-preview": {
662
- "id": "qwen3.6-max-preview",
663
- "name": "Qwen3.6 Max Preview",
664
- "description": "Flagship Qwen model for complex reasoning, coding, and agentic workflows",
824
+ "qwen-vl-ocr": {
825
+ "id": "qwen-vl-ocr",
826
+ "name": "Qwen-VL OCR",
827
+ "description": "OCR model for extracting structured text from documents and screenshots",
665
828
  "family": "qwen",
666
829
  "attachment": false,
667
- "reasoning": true,
668
- "reasoning_options": [
669
- {
670
- "type": "toggle"
671
- },
672
- {
673
- "type": "budget_tokens"
674
- }
675
- ],
676
- "tool_call": true,
830
+ "reasoning": false,
831
+ "tool_call": false,
677
832
  "temperature": true,
678
- "knowledge": "2025-04",
679
- "release_date": "2026-04-20",
680
- "last_updated": "2026-04-20",
833
+ "knowledge": "2024-04",
834
+ "release_date": "2024-10-28",
835
+ "last_updated": "2025-04-13",
681
836
  "modalities": {
682
837
  "input": [
683
- "text"
838
+ "text",
839
+ "image"
684
840
  ],
685
841
  "output": [
686
842
  "text"
@@ -688,97 +844,69 @@
688
844
  },
689
845
  "open_weights": false,
690
846
  "limit": {
691
- "context": 262144,
692
- "output": 65536
847
+ "context": 34096,
848
+ "output": 4096
693
849
  },
694
850
  "cost": {
695
- "input": 1.3,
696
- "output": 7.8,
697
- "cache_read": 0.13,
698
- "cache_write": 1.625
851
+ "input": 0.72,
852
+ "output": 0.72
699
853
  }
700
854
  },
701
- "qwen3.8-max": {
702
- "id": "qwen3.8-max",
703
- "name": "Qwen3.8 Max",
704
- "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
855
+ "qwen3-32b": {
856
+ "id": "qwen3-32b",
857
+ "name": "Qwen3 32B",
858
+ "description": "Dense open Qwen model for self-hosted chat, reasoning, and coding",
705
859
  "family": "qwen",
706
- "attachment": true,
860
+ "attachment": false,
707
861
  "reasoning": true,
708
862
  "reasoning_options": [
709
863
  {
710
864
  "type": "toggle"
711
865
  },
712
866
  {
713
- "type": "effort",
714
- "values": [
715
- "low",
716
- "medium",
717
- "xhigh"
718
- ]
719
- },
720
- {
721
- "type": "budget_tokens",
722
- "min": 0,
723
- "max": 262144
867
+ "type": "budget_tokens"
724
868
  }
725
869
  ],
726
870
  "tool_call": true,
727
- "interleaved": {
728
- "field": "reasoning_content"
729
- },
730
- "structured_output": true,
731
871
  "temperature": true,
732
- "release_date": "2026-08-03",
733
- "last_updated": "2026-08-03",
872
+ "knowledge": "2025-04",
873
+ "release_date": "2025-04",
874
+ "last_updated": "2025-04",
734
875
  "modalities": {
735
876
  "input": [
736
- "text",
737
- "image",
738
- "video",
739
- "pdf"
877
+ "text"
740
878
  ],
741
879
  "output": [
742
880
  "text"
743
881
  ]
744
882
  },
745
- "open_weights": false,
883
+ "open_weights": true,
746
884
  "limit": {
747
- "context": 1000000,
748
- "output": 131072
885
+ "context": 131072,
886
+ "output": 16384
749
887
  },
750
888
  "cost": {
751
- "input": 2,
752
- "output": 6,
753
- "cache_read": 0.25,
754
- "cache_write": 2.5
889
+ "input": 0.7,
890
+ "output": 2.8,
891
+ "reasoning": 8.4
755
892
  }
756
893
  },
757
- "qwen3.7-plus": {
758
- "id": "qwen3.7-plus",
759
- "name": "Qwen3.7 Plus",
760
- "description": "Multimodal Qwen workhorse for long-context agents, visual inputs, and coding",
894
+ "qwq-plus": {
895
+ "id": "qwq-plus",
896
+ "name": "QwQ Plus",
897
+ "description": "Qwen reasoning model for deliberate problem solving, math, and coding",
761
898
  "family": "qwen",
762
899
  "attachment": false,
763
900
  "reasoning": true,
764
- "reasoning_options": [
765
- {
766
- "type": "toggle"
767
- },
768
- {
769
- "type": "budget_tokens"
770
- }
771
- ],
901
+ "reasoning_options": [],
772
902
  "tool_call": true,
773
903
  "temperature": true,
774
- "knowledge": "2025-04",
775
- "release_date": "2026-06-02",
776
- "last_updated": "2026-06-04",
904
+ "knowledge": "2024-04",
905
+ "release_date": "2025-03-05",
906
+ "last_updated": "2025-03-05",
777
907
  "modalities": {
778
908
  "input": [
779
- "text",
780
- "image",
781
- "video"
909
+ "text"
782
910
  ],
783
911
  "output": [
784
912
  "text"
@@ -786,42 +914,25 @@
786
914
  },
787
915
  "open_weights": false,
788
916
  "limit": {
789
- "context": 1000000,
790
- "output": 65536
917
+ "context": 131072,
918
+ "output": 8192
791
919
  },
792
920
  "cost": {
793
- "input": 0.5,
794
- "output": 3,
795
- "cache_read": 0.05,
796
- "cache_write": 0.625,
797
- "tiers": [
798
- {
799
- "input": 2,
800
- "output": 6,
801
- "cache_read": 0.2,
802
- "cache_write": 2.5,
803
- "tier": {
804
- "type": "context",
805
- "size": 256000
806
- }
807
- }
808
- ],
809
- "context_over_200k": {
810
- "input": 2,
811
- "output": 6,
812
- "cache_read": 0.2,
813
- "cache_write": 2.5
814
- }
921
+ "input": 0.8,
922
+ "output": 2.4
815
923
  }
816
924
  },
817
- "qwen3-vl-235b-a22b": {
818
- "id": "qwen3-vl-235b-a22b",
819
- "name": "Qwen3-VL 235B-A22B",
925
+ "qwen3-vl-plus": {
926
+ "id": "qwen3-vl-plus",
927
+ "name": "Qwen3-VL Plus",
820
928
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
821
929
  "family": "qwen",
822
930
  "attachment": false,
823
931
  "reasoning": true,
824
932
  "reasoning_options": [
933
+ {
934
+ "type": "toggle"
935
+ },
825
936
  {
826
937
  "type": "budget_tokens"
827
938
  }
@@ -829,8 +940,8 @@
829
940
  "tool_call": true,
830
941
  "temperature": true,
831
942
  "knowledge": "2025-04",
832
- "release_date": "2025-04",
833
- "last_updated": "2025-04",
943
+ "release_date": "2025-09-23",
944
+ "last_updated": "2025-09-23",
834
945
  "modalities": {
835
946
  "input": [
836
947
  "text",
@@ -840,60 +951,59 @@
840
951
  "text"
841
952
  ]
842
953
  },
843
- "open_weights": true,
954
+ "open_weights": false,
844
955
  "limit": {
845
- "context": 131072,
956
+ "context": 262144,
846
957
  "output": 32768
847
958
  },
848
959
  "cost": {
849
- "input": 0.7,
850
- "output": 2.8,
851
- "reasoning": 8.4
960
+ "input": 0.2,
961
+ "output": 1.6,
962
+ "reasoning": 4.8
852
963
  }
853
964
  },
854
- "qwen-vl-ocr": {
855
- "id": "qwen-vl-ocr",
856
- "name": "Qwen-VL OCR",
857
- "description": "OCR model for extracting structured text from documents and screenshots",
965
+ "qwen2-5-7b-instruct": {
966
+ "id": "qwen2-5-7b-instruct",
967
+ "name": "Qwen2.5 7B Instruct",
968
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
858
969
  "family": "qwen",
859
970
  "attachment": false,
860
971
  "reasoning": false,
861
- "tool_call": false,
972
+ "tool_call": true,
862
973
  "temperature": true,
863
974
  "knowledge": "2024-04",
864
- "release_date": "2024-10-28",
865
- "last_updated": "2025-04-13",
975
+ "release_date": "2024-09",
976
+ "last_updated": "2024-09",
866
977
  "modalities": {
867
978
  "input": [
868
- "text",
869
- "image"
979
+ "text"
870
980
  ],
871
981
  "output": [
872
982
  "text"
873
983
  ]
874
984
  },
875
- "open_weights": false,
985
+ "open_weights": true,
876
986
  "limit": {
877
- "context": 34096,
878
- "output": 4096
987
+ "context": 131072,
988
+ "output": 8192
879
989
  },
880
990
  "cost": {
881
- "input": 0.72,
882
- "output": 0.72
991
+ "input": 0.175,
992
+ "output": 0.7
883
993
  }
884
994
  },
885
- "qwen2-5-72b-instruct": {
886
- "id": "qwen2-5-72b-instruct",
887
- "name": "Qwen2.5 72B Instruct",
888
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
995
+ "qwen3-coder-30b-a3b-instruct": {
996
+ "id": "qwen3-coder-30b-a3b-instruct",
997
+ "name": "Qwen3-Coder 30B-A3B Instruct",
998
+ "description": "Smaller Qwen coder for efficient local agents and repo-level fixes",
889
999
  "family": "qwen",
890
1000
  "attachment": false,
891
1001
  "reasoning": false,
892
1002
  "tool_call": true,
893
1003
  "temperature": true,
894
- "knowledge": "2024-04",
895
- "release_date": "2024-09",
896
- "last_updated": "2024-09",
1004
+ "knowledge": "2025-04",
1005
+ "release_date": "2025-04",
1006
+ "last_updated": "2025-04",
897
1007
  "modalities": {
898
1008
  "input": [
899
1009
  "text"
@@ -904,20 +1014,38 @@
904
1014
  },
905
1015
  "open_weights": true,
906
1016
  "limit": {
907
- "context": 131072,
908
- "output": 8192
1017
+ "context": 262144,
1018
+ "output": 65536
909
1019
  },
910
1020
  "cost": {
911
- "input": 1.4,
912
- "output": 5.6
1021
+ "input": 0.45,
1022
+ "output": 2.25,
1023
+ "tiers": [
1024
+ {
1025
+ "input": 0.75,
1026
+ "output": 3.75,
1027
+ "tier": {
1028
+ "type": "context",
1029
+ "size": 32000
1030
+ }
1031
+ },
1032
+ {
1033
+ "input": 1.2,
1034
+ "output": 6,
1035
+ "tier": {
1036
+ "type": "context",
1037
+ "size": 128000
1038
+ }
1039
+ }
1040
+ ]
913
1041
  }
914
1042
  },
915
- "qwen3-14b": {
916
- "id": "qwen3-14b",
917
- "name": "Qwen3 14B",
918
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1043
+ "qwen3.5-397b-a17b": {
1044
+ "id": "qwen3.5-397b-a17b",
1045
+ "name": "Qwen3.5 397B-A17B",
1046
+ "description": "Large open Qwen multimodal MoE for visual agents and long technical tasks",
919
1047
  "family": "qwen",
920
- "attachment": false,
1048
+ "attachment": true,
921
1049
  "reasoning": true,
922
1050
  "reasoning_options": [
923
1051
  {
@@ -928,13 +1056,16 @@
928
1056
  }
929
1057
  ],
930
1058
  "tool_call": true,
1059
+ "structured_output": true,
931
1060
  "temperature": true,
932
- "knowledge": "2025-04",
933
- "release_date": "2025-04",
934
- "last_updated": "2025-04",
1061
+ "release_date": "2026-02-15",
1062
+ "last_updated": "2026-02-15",
935
1063
  "modalities": {
936
1064
  "input": [
937
- "text"
1065
+ "text",
1066
+ "image",
1067
+ "video",
1068
+ "audio"
938
1069
  ],
939
1070
  "output": [
940
1071
  "text"
@@ -942,65 +1073,40 @@
942
1073
  },
943
1074
  "open_weights": true,
944
1075
  "limit": {
945
- "context": 131072,
946
- "output": 8192
947
- },
948
- "cost": {
949
- "input": 0.35,
950
- "output": 1.4,
951
- "reasoning": 4.2
952
- }
953
- },
954
- "qwen-mt-turbo": {
955
- "id": "qwen-mt-turbo",
956
- "name": "Qwen-MT Turbo",
957
- "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
958
- "family": "qwen",
959
- "attachment": false,
960
- "reasoning": false,
961
- "tool_call": false,
962
- "temperature": true,
963
- "knowledge": "2024-04",
964
- "release_date": "2025-01",
965
- "last_updated": "2025-01",
966
- "modalities": {
967
- "input": [
968
- "text"
969
- ],
970
- "output": [
971
- "text"
972
- ]
973
- },
974
- "open_weights": false,
975
- "limit": {
976
- "context": 16384,
977
- "output": 8192
1076
+ "context": 262144,
1077
+ "output": 65536
978
1078
  },
979
1079
  "cost": {
980
- "input": 0.16,
981
- "output": 0.49
1080
+ "input": 0.6,
1081
+ "output": 3.6
982
1082
  }
983
1083
  },
984
- "qwen3-next-80b-a3b-thinking": {
985
- "id": "qwen3-next-80b-a3b-thinking",
986
- "name": "Qwen3-Next 80B-A3B (Thinking)",
987
- "description": "Efficient Qwen thinking model for local reasoning, math, and coding agents",
1084
+ "qwen3.6-27b": {
1085
+ "id": "qwen3.6-27b",
1086
+ "name": "Qwen3.6 27B",
1087
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
988
1088
  "family": "qwen",
989
- "attachment": false,
1089
+ "attachment": true,
990
1090
  "reasoning": true,
991
1091
  "reasoning_options": [
1092
+ {
1093
+ "type": "toggle"
1094
+ },
992
1095
  {
993
1096
  "type": "budget_tokens"
994
1097
  }
995
1098
  ],
996
1099
  "tool_call": true,
1100
+ "structured_output": true,
997
1101
  "temperature": true,
998
- "knowledge": "2025-04",
999
- "release_date": "2025-09",
1000
- "last_updated": "2025-09",
1102
+ "release_date": "2026-04-22",
1103
+ "last_updated": "2026-04-22",
1001
1104
  "modalities": {
1002
1105
  "input": [
1003
- "text"
1106
+ "text",
1107
+ "image",
1108
+ "video",
1109
+ "audio"
1004
1110
  ],
1005
1111
  "output": [
1006
1112
  "text"
@@ -1008,42 +1114,42 @@
1008
1114
  },
1009
1115
  "open_weights": true,
1010
1116
  "limit": {
1011
- "context": 131072,
1012
- "output": 32768
1117
+ "context": 262144,
1118
+ "output": 65536
1013
1119
  },
1014
1120
  "cost": {
1015
- "input": 0.5,
1016
- "output": 6
1121
+ "input": 0.6,
1122
+ "output": 3.6
1017
1123
  }
1018
1124
  },
1019
- "qwen3-coder-480b-a35b-instruct": {
1020
- "id": "qwen3-coder-480b-a35b-instruct",
1021
- "name": "Qwen3-Coder 480B-A35B Instruct",
1022
- "description": "Open Qwen coding heavyweight for repository reasoning and agentic engineering",
1125
+ "qwen3-asr-flash": {
1126
+ "id": "qwen3-asr-flash",
1127
+ "name": "Qwen3-ASR Flash",
1128
+ "description": "Speech transcription model for accurate audio-to-text and captioning workflows",
1023
1129
  "family": "qwen",
1024
1130
  "attachment": false,
1025
1131
  "reasoning": false,
1026
- "tool_call": true,
1027
- "temperature": true,
1028
- "knowledge": "2025-04",
1029
- "release_date": "2025-04",
1030
- "last_updated": "2025-04",
1132
+ "tool_call": false,
1133
+ "temperature": false,
1134
+ "knowledge": "2024-04",
1135
+ "release_date": "2025-09-08",
1136
+ "last_updated": "2025-09-08",
1031
1137
  "modalities": {
1032
1138
  "input": [
1033
- "text"
1139
+ "audio"
1034
1140
  ],
1035
1141
  "output": [
1036
1142
  "text"
1037
1143
  ]
1038
1144
  },
1039
- "open_weights": true,
1145
+ "open_weights": false,
1040
1146
  "limit": {
1041
- "context": 262144,
1042
- "output": 65536
1147
+ "context": 53248,
1148
+ "output": 4096
1043
1149
  },
1044
1150
  "cost": {
1045
- "input": 1.5,
1046
- "output": 7.5
1151
+ "input": 0.035,
1152
+ "output": 0.035
1047
1153
  }
1048
1154
  },
1049
1155
  "qwen3-omni-flash": {
@@ -1087,50 +1193,43 @@
1087
1193
  "output_audio": 15.11
1088
1194
  }
1089
1195
  },
1090
- "qwen-flash": {
1091
- "id": "qwen-flash",
1092
- "name": "Qwen Flash",
1093
- "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
1196
+ "qwen2-5-vl-72b-instruct": {
1197
+ "id": "qwen2-5-vl-72b-instruct",
1198
+ "name": "Qwen2.5-VL 72B Instruct",
1199
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1094
1200
  "family": "qwen",
1095
1201
  "attachment": false,
1096
- "reasoning": true,
1097
- "reasoning_options": [
1098
- {
1099
- "type": "toggle"
1100
- },
1101
- {
1102
- "type": "budget_tokens"
1103
- }
1104
- ],
1202
+ "reasoning": false,
1105
1203
  "tool_call": true,
1106
1204
  "temperature": true,
1107
1205
  "knowledge": "2024-04",
1108
- "release_date": "2025-07-28",
1109
- "last_updated": "2025-07-28",
1206
+ "release_date": "2024-09",
1207
+ "last_updated": "2024-09",
1110
1208
  "modalities": {
1111
1209
  "input": [
1112
- "text"
1210
+ "text",
1211
+ "image"
1113
1212
  ],
1114
1213
  "output": [
1115
1214
  "text"
1116
1215
  ]
1117
1216
  },
1118
- "open_weights": false,
1217
+ "open_weights": true,
1119
1218
  "limit": {
1120
- "context": 1000000,
1121
- "output": 32768
1219
+ "context": 131072,
1220
+ "output": 8192
1122
1221
  },
1123
1222
  "cost": {
1124
- "input": 0.05,
1125
- "output": 0.4
1223
+ "input": 2.8,
1224
+ "output": 8.4
1126
1225
  }
1127
1226
  },
1128
- "qwen3-8b": {
1129
- "id": "qwen3-8b",
1130
- "name": "Qwen3 8B",
1131
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1227
+ "qwen3.6-35b-a3b": {
1228
+ "id": "qwen3.6-35b-a3b",
1229
+ "name": "Qwen3.6 35B-A3B",
1230
+ "description": "Open multimodal Qwen MoE for local agents that need vision, audio, and code",
1132
1231
  "family": "qwen",
1133
- "attachment": false,
1232
+ "attachment": true,
1134
1233
  "reasoning": true,
1135
1234
  "reasoning_options": [
1136
1235
  {
@@ -1141,13 +1240,16 @@
1141
1240
  }
1142
1241
  ],
1143
1242
  "tool_call": true,
1243
+ "structured_output": true,
1144
1244
  "temperature": true,
1145
- "knowledge": "2025-04",
1146
- "release_date": "2025-04",
1147
- "last_updated": "2025-04",
1245
+ "release_date": "2026-04-17",
1246
+ "last_updated": "2026-04-17",
1148
1247
  "modalities": {
1149
1248
  "input": [
1150
- "text"
1249
+ "text",
1250
+ "image",
1251
+ "video",
1252
+ "audio"
1151
1253
  ],
1152
1254
  "output": [
1153
1255
  "text"
@@ -1155,13 +1257,12 @@
1155
1257
  },
1156
1258
  "open_weights": true,
1157
1259
  "limit": {
1158
- "context": 131072,
1159
- "output": 8192
1260
+ "context": 262144,
1261
+ "output": 65536
1160
1262
  },
1161
1263
  "cost": {
1162
- "input": 0.18,
1163
- "output": 0.7,
1164
- "reasoning": 2.1
1264
+ "input": 0.248,
1265
+ "output": 1.485
1165
1266
  }
1166
1267
  },
1167
1268
  "qwen3-max": {
@@ -1194,35 +1295,26 @@
1194
1295
  "output": 6
1195
1296
  }
1196
1297
  },
1197
- "glm-5.2": {
1198
- "id": "glm-5.2",
1199
- "name": "GLM-5.2",
1200
- "description": "Open flagship GLM for long-horizon coding agents and million-token context work",
1201
- "family": "glm",
1298
+ "qwen-plus": {
1299
+ "id": "qwen-plus",
1300
+ "name": "Qwen Plus",
1301
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1302
+ "family": "qwen",
1202
1303
  "attachment": false,
1203
1304
  "reasoning": true,
1204
1305
  "reasoning_options": [
1205
1306
  {
1206
- "type": "effort",
1207
- "values": [
1208
- "none",
1209
- "minimal",
1210
- "low",
1211
- "medium",
1212
- "high",
1213
- "xhigh",
1214
- "max"
1215
- ]
1307
+ "type": "toggle"
1308
+ },
1309
+ {
1310
+ "type": "budget_tokens"
1216
1311
  }
1217
1312
  ],
1218
1313
  "tool_call": true,
1219
- "interleaved": {
1220
- "field": "reasoning_content"
1221
- },
1222
- "structured_output": true,
1223
1314
  "temperature": true,
1224
- "release_date": "2026-06-13",
1225
- "last_updated": "2026-06-13",
1315
+ "knowledge": "2024-04",
1316
+ "release_date": "2024-01-25",
1317
+ "last_updated": "2025-09-11",
1226
1318
  "modalities": {
1227
1319
  "input": [
1228
1320
  "text"
@@ -1231,57 +1323,21 @@
1231
1323
  "text"
1232
1324
  ]
1233
1325
  },
1234
- "open_weights": true,
1326
+ "open_weights": false,
1235
1327
  "limit": {
1236
1328
  "context": 1000000,
1237
- "output": 131072
1238
- },
1239
- "cost": {
1240
- "input": 1.4,
1241
- "output": 4.4,
1242
- "cache_read": 0.28,
1243
- "cache_write": 0
1244
- }
1245
- },
1246
- "qwen2-5-omni-7b": {
1247
- "id": "qwen2-5-omni-7b",
1248
- "name": "Qwen2.5-Omni 7B",
1249
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1250
- "family": "qwen",
1251
- "attachment": false,
1252
- "reasoning": false,
1253
- "tool_call": true,
1254
- "temperature": true,
1255
- "knowledge": "2024-04",
1256
- "release_date": "2024-12",
1257
- "last_updated": "2024-12",
1258
- "modalities": {
1259
- "input": [
1260
- "text",
1261
- "image",
1262
- "audio",
1263
- "video"
1264
- ],
1265
- "output": [
1266
- "text",
1267
- "audio"
1268
- ]
1269
- },
1270
- "open_weights": true,
1271
- "limit": {
1272
- "context": 32768,
1273
- "output": 2048
1329
+ "output": 32768
1274
1330
  },
1275
1331
  "cost": {
1276
- "input": 0.1,
1277
- "output": 0.4,
1278
- "input_audio": 6.76
1332
+ "input": 0.4,
1333
+ "output": 1.2,
1334
+ "reasoning": 4
1279
1335
  }
1280
1336
  },
1281
- "qwen3.6-35b-a3b": {
1282
- "id": "qwen3.6-35b-a3b",
1283
- "name": "Qwen3.6 35B-A3B",
1284
- "description": "Open multimodal Qwen MoE for local agents that need vision, audio, and code",
1337
+ "qwen3.5-122b-a10b": {
1338
+ "id": "qwen3.5-122b-a10b",
1339
+ "name": "Qwen3.5 122B-A10B",
1340
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1285
1341
  "family": "qwen",
1286
1342
  "attachment": true,
1287
1343
  "reasoning": true,
@@ -1296,8 +1352,8 @@
1296
1352
  "tool_call": true,
1297
1353
  "structured_output": true,
1298
1354
  "temperature": true,
1299
- "release_date": "2026-04-17",
1300
- "last_updated": "2026-04-17",
1355
+ "release_date": "2026-02-23",
1356
+ "last_updated": "2026-02-23",
1301
1357
  "modalities": {
1302
1358
  "input": [
1303
1359
  "text",
@@ -1315,98 +1371,62 @@
1315
1371
  "output": 65536
1316
1372
  },
1317
1373
  "cost": {
1318
- "input": 0.248,
1319
- "output": 1.485
1374
+ "input": 0.4,
1375
+ "output": 3.2
1320
1376
  }
1321
1377
  },
1322
- "qvq-max": {
1323
- "id": "qvq-max",
1324
- "name": "QVQ Max",
1325
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1326
- "family": "qvq",
1378
+ "qwen3-omni-flash-realtime": {
1379
+ "id": "qwen3-omni-flash-realtime",
1380
+ "name": "Qwen3-Omni Flash Realtime",
1381
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1382
+ "family": "qwen",
1327
1383
  "attachment": false,
1328
- "reasoning": true,
1329
- "reasoning_options": [],
1384
+ "reasoning": false,
1330
1385
  "tool_call": true,
1331
1386
  "temperature": true,
1332
1387
  "knowledge": "2024-04",
1333
- "release_date": "2025-03-25",
1334
- "last_updated": "2025-03-25",
1388
+ "release_date": "2025-09-15",
1389
+ "last_updated": "2025-09-15",
1335
1390
  "modalities": {
1336
1391
  "input": [
1337
1392
  "text",
1338
- "image"
1393
+ "image",
1394
+ "audio",
1395
+ "video"
1339
1396
  ],
1340
1397
  "output": [
1341
- "text"
1342
- ]
1343
- },
1344
- "open_weights": false,
1345
- "limit": {
1346
- "context": 131072,
1347
- "output": 8192
1348
- },
1349
- "cost": {
1350
- "input": 1.2,
1351
- "output": 4.8
1352
- }
1353
- },
1354
- "qwen3.6-27b": {
1355
- "id": "qwen3.6-27b",
1356
- "name": "Qwen3.6 27B",
1357
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1358
- "family": "qwen",
1359
- "attachment": true,
1360
- "reasoning": true,
1361
- "reasoning_options": [
1362
- {
1363
- "type": "toggle"
1364
- },
1365
- {
1366
- "type": "budget_tokens"
1367
- }
1368
- ],
1369
- "tool_call": true,
1370
- "structured_output": true,
1371
- "temperature": true,
1372
- "release_date": "2026-04-22",
1373
- "last_updated": "2026-04-22",
1374
- "modalities": {
1375
- "input": [
1376
1398
  "text",
1377
- "image",
1378
- "video",
1379
1399
  "audio"
1380
- ],
1381
- "output": [
1382
- "text"
1383
1400
  ]
1384
1401
  },
1385
- "open_weights": true,
1402
+ "open_weights": false,
1386
1403
  "limit": {
1387
- "context": 262144,
1388
- "output": 65536
1404
+ "context": 65536,
1405
+ "output": 16384
1389
1406
  },
1390
1407
  "cost": {
1391
- "input": 0.6,
1392
- "output": 3.6
1408
+ "input": 0.52,
1409
+ "output": 1.99,
1410
+ "input_audio": 4.57,
1411
+ "output_audio": 18.13
1393
1412
  }
1394
1413
  },
1395
- "qwen3-next-80b-a3b-instruct": {
1396
- "id": "qwen3-next-80b-a3b-instruct",
1397
- "name": "Qwen3-Next 80B-A3B Instruct",
1398
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1414
+ "qwen2-5-vl-7b-instruct": {
1415
+ "id": "qwen2-5-vl-7b-instruct",
1416
+ "name": "Qwen2.5-VL 7B Instruct",
1417
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1399
1418
  "family": "qwen",
1400
1419
  "attachment": false,
1401
1420
  "reasoning": false,
1402
1421
  "tool_call": true,
1403
1422
  "temperature": true,
1404
- "knowledge": "2025-04",
1405
- "release_date": "2025-09",
1406
- "last_updated": "2025-09",
1423
+ "knowledge": "2024-04",
1424
+ "release_date": "2024-09",
1425
+ "last_updated": "2024-09",
1407
1426
  "modalities": {
1408
1427
  "input": [
1409
- "text"
1428
+ "text",
1429
+ "image"
1410
1430
  ],
1411
1431
  "output": [
1412
1432
  "text"
@@ -1415,69 +1435,85 @@
1415
1435
  "open_weights": true,
1416
1436
  "limit": {
1417
1437
  "context": 131072,
1418
- "output": 32768
1438
+ "output": 8192
1419
1439
  },
1420
1440
  "cost": {
1421
- "input": 0.5,
1422
- "output": 2
1441
+ "input": 0.35,
1442
+ "output": 1.05
1423
1443
  }
1424
1444
  },
1425
- "qwen3-omni-flash-realtime": {
1426
- "id": "qwen3-omni-flash-realtime",
1427
- "name": "Qwen3-Omni Flash Realtime",
1428
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1429
- "family": "qwen",
1430
- "attachment": false,
1431
- "reasoning": false,
1445
+ "qwen3.6-flash": {
1446
+ "id": "qwen3.6-flash",
1447
+ "name": "Qwen3.6 Flash",
1448
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1449
+ "family": "qwen3.6",
1450
+ "attachment": true,
1451
+ "reasoning": true,
1452
+ "reasoning_options": [
1453
+ {
1454
+ "type": "toggle"
1455
+ },
1456
+ {
1457
+ "type": "budget_tokens"
1458
+ }
1459
+ ],
1432
1460
  "tool_call": true,
1461
+ "structured_output": true,
1433
1462
  "temperature": true,
1434
- "knowledge": "2024-04",
1435
- "release_date": "2025-09-15",
1436
- "last_updated": "2025-09-15",
1463
+ "release_date": "2026-04-27",
1464
+ "last_updated": "2026-04-27",
1437
1465
  "modalities": {
1438
1466
  "input": [
1439
1467
  "text",
1440
1468
  "image",
1441
- "audio",
1442
1469
  "video"
1443
1470
  ],
1444
1471
  "output": [
1445
- "text",
1446
- "audio"
1472
+ "text"
1447
1473
  ]
1448
1474
  },
1449
1475
  "open_weights": false,
1450
1476
  "limit": {
1451
- "context": 65536,
1452
- "output": 16384
1477
+ "context": 1000000,
1478
+ "output": 65536
1453
1479
  },
1454
1480
  "cost": {
1455
- "input": 0.52,
1456
- "output": 1.99,
1457
- "input_audio": 4.57,
1458
- "output_audio": 18.13
1481
+ "input": 0.1875,
1482
+ "output": 1.125,
1483
+ "cache_write": 0.234375
1459
1484
  }
1460
1485
  },
1461
- "qwen3.6-plus": {
1462
- "id": "qwen3.6-plus",
1463
- "name": "Qwen3.6 Plus",
1464
- "description": "Earlier Qwen multimodal workhorse for million-token agent and document tasks",
1486
+ "qwen3.8-flash": {
1487
+ "id": "qwen3.8-flash",
1488
+ "name": "Qwen3.8 Flash",
1489
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1465
1490
  "family": "qwen",
1466
- "attachment": false,
1491
+ "attachment": true,
1467
1492
  "reasoning": true,
1468
1493
  "reasoning_options": [
1469
1494
  {
1470
1495
  "type": "toggle"
1471
1496
  },
1472
1497
  {
1473
- "type": "budget_tokens"
1498
+ "type": "effort",
1499
+ "values": [
1500
+ "low",
1501
+ "medium",
1502
+ "xhigh"
1503
+ ]
1504
+ },
1505
+ {
1506
+ "type": "budget_tokens",
1507
+ "max": 262144
1474
1508
  }
1475
1509
  ],
1476
1510
  "tool_call": true,
1477
- "temperature": true,
1478
- "knowledge": "2025-04",
1479
- "release_date": "2026-04-02",
1480
- "last_updated": "2026-04-02",
1511
+ "interleaved": {
1512
+ "field": "reasoning_content"
1513
+ },
1514
+ "structured_output": true,
1515
+ "release_date": "2026-08-26",
1516
+ "last_updated": "2026-08-26",
1481
1517
  "modalities": {
1482
1518
  "input": [
1483
1519
  "text",
@@ -1491,31 +1527,13 @@
1491
1527
  "open_weights": false,
1492
1528
  "limit": {
1493
1529
  "context": 1000000,
1494
- "output": 65536
1530
+ "output": 131072
1495
1531
  },
1496
1532
  "cost": {
1497
- "input": 0.5,
1498
- "output": 3,
1499
- "cache_read": 0.05,
1500
- "cache_write": 0.625,
1501
- "tiers": [
1502
- {
1503
- "input": 2,
1504
- "output": 6,
1505
- "cache_read": 0.2,
1506
- "cache_write": 2.5,
1507
- "tier": {
1508
- "type": "context",
1509
- "size": 256000
1510
- }
1511
- }
1512
- ],
1513
- "context_over_200k": {
1514
- "input": 2,
1515
- "output": 6,
1516
- "cache_read": 0.2,
1517
- "cache_write": 2.5
1518
- }
1533
+ "input": 0.15,
1534
+ "output": 0.47,
1535
+ "cache_read": 0.016,
1536
+ "cache_write": 0.2
1519
1537
  }
1520
1538
  },
1521
1539
  "qwen2-5-14b-instruct": {
@@ -1548,18 +1566,26 @@
1548
1566
  "output": 1.4
1549
1567
  }
1550
1568
  },
1551
- "qwen-mt-plus": {
1552
- "id": "qwen-mt-plus",
1553
- "name": "Qwen-MT Plus",
1554
- "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
1555
- "family": "qwen",
1556
- "attachment": false,
1557
- "reasoning": false,
1558
- "tool_call": false,
1569
+ "qwen3.6-max-preview": {
1570
+ "id": "qwen3.6-max-preview",
1571
+ "name": "Qwen3.6 Max Preview",
1572
+ "description": "Flagship Qwen model for complex reasoning, coding, and agentic workflows",
1573
+ "family": "qwen",
1574
+ "attachment": false,
1575
+ "reasoning": true,
1576
+ "reasoning_options": [
1577
+ {
1578
+ "type": "toggle"
1579
+ },
1580
+ {
1581
+ "type": "budget_tokens"
1582
+ }
1583
+ ],
1584
+ "tool_call": true,
1559
1585
  "temperature": true,
1560
- "knowledge": "2024-04",
1561
- "release_date": "2025-01",
1562
- "last_updated": "2025-01",
1586
+ "knowledge": "2025-04",
1587
+ "release_date": "2026-04-20",
1588
+ "last_updated": "2026-04-20",
1563
1589
  "modalities": {
1564
1590
  "input": [
1565
1591
  "text"
@@ -1570,12 +1596,14 @@
1570
1596
  },
1571
1597
  "open_weights": false,
1572
1598
  "limit": {
1573
- "context": 16384,
1574
- "output": 8192
1599
+ "context": 262144,
1600
+ "output": 65536
1575
1601
  },
1576
1602
  "cost": {
1577
- "input": 2.46,
1578
- "output": 7.37
1603
+ "input": 1.3,
1604
+ "output": 7.8,
1605
+ "cache_read": 0.13,
1606
+ "cache_write": 1.625
1579
1607
  }
1580
1608
  },
1581
1609
  "qwen-plus-character-ja": {
@@ -1608,40 +1636,10 @@
1608
1636
  "output": 1.4
1609
1637
  }
1610
1638
  },
1611
- "qwen2-5-7b-instruct": {
1612
- "id": "qwen2-5-7b-instruct",
1613
- "name": "Qwen2.5 7B Instruct",
1614
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1615
- "family": "qwen",
1616
- "attachment": false,
1617
- "reasoning": false,
1618
- "tool_call": true,
1619
- "temperature": true,
1620
- "knowledge": "2024-04",
1621
- "release_date": "2024-09",
1622
- "last_updated": "2024-09",
1623
- "modalities": {
1624
- "input": [
1625
- "text"
1626
- ],
1627
- "output": [
1628
- "text"
1629
- ]
1630
- },
1631
- "open_weights": true,
1632
- "limit": {
1633
- "context": 131072,
1634
- "output": 8192
1635
- },
1636
- "cost": {
1637
- "input": 0.175,
1638
- "output": 0.7
1639
- }
1640
- },
1641
- "qwen3.5-27b": {
1642
- "id": "qwen3.5-27b",
1643
- "name": "Qwen3.5 27B",
1644
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1639
+ "qwen3.8-max": {
1640
+ "id": "qwen3.8-max",
1641
+ "name": "Qwen3.8 Max",
1642
+ "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
1645
1643
  "family": "qwen",
1646
1644
  "attachment": true,
1647
1645
  "reasoning": true,
@@ -1650,48 +1648,70 @@
1650
1648
  "type": "toggle"
1651
1649
  },
1652
1650
  {
1653
- "type": "budget_tokens"
1651
+ "type": "effort",
1652
+ "values": [
1653
+ "low",
1654
+ "medium",
1655
+ "xhigh"
1656
+ ]
1657
+ },
1658
+ {
1659
+ "type": "budget_tokens",
1660
+ "min": 0,
1661
+ "max": 262144
1654
1662
  }
1655
1663
  ],
1656
1664
  "tool_call": true,
1665
+ "interleaved": {
1666
+ "field": "reasoning_content"
1667
+ },
1657
1668
  "structured_output": true,
1658
1669
  "temperature": true,
1659
- "release_date": "2026-02-23",
1660
- "last_updated": "2026-02-23",
1670
+ "release_date": "2026-08-03",
1671
+ "last_updated": "2026-08-03",
1661
1672
  "modalities": {
1662
1673
  "input": [
1663
1674
  "text",
1664
1675
  "image",
1665
1676
  "video",
1666
- "audio"
1677
+ "pdf"
1667
1678
  ],
1668
1679
  "output": [
1669
1680
  "text"
1670
1681
  ]
1671
1682
  },
1672
- "open_weights": true,
1683
+ "open_weights": false,
1673
1684
  "limit": {
1674
- "context": 262144,
1675
- "output": 65536
1685
+ "context": 1000000,
1686
+ "output": 131072
1676
1687
  },
1677
1688
  "cost": {
1678
- "input": 0.3,
1679
- "output": 2.4
1689
+ "input": 2,
1690
+ "output": 6,
1691
+ "cache_read": 0.25,
1692
+ "cache_write": 2.5
1680
1693
  }
1681
1694
  },
1682
- "qwq-plus": {
1683
- "id": "qwq-plus",
1684
- "name": "QwQ Plus",
1685
- "description": "Qwen reasoning model for deliberate problem solving, math, and coding",
1695
+ "qwen3-8b": {
1696
+ "id": "qwen3-8b",
1697
+ "name": "Qwen3 8B",
1698
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1686
1699
  "family": "qwen",
1687
1700
  "attachment": false,
1688
1701
  "reasoning": true,
1689
- "reasoning_options": [],
1702
+ "reasoning_options": [
1703
+ {
1704
+ "type": "toggle"
1705
+ },
1706
+ {
1707
+ "type": "budget_tokens"
1708
+ }
1709
+ ],
1690
1710
  "tool_call": true,
1691
1711
  "temperature": true,
1692
- "knowledge": "2024-04",
1693
- "release_date": "2025-03-05",
1694
- "last_updated": "2025-03-05",
1712
+ "knowledge": "2025-04",
1713
+ "release_date": "2025-04",
1714
+ "last_updated": "2025-04",
1695
1715
  "modalities": {
1696
1716
  "input": [
1697
1717
  "text"
@@ -1700,22 +1720,23 @@
1700
1720
  "text"
1701
1721
  ]
1702
1722
  },
1703
- "open_weights": false,
1723
+ "open_weights": true,
1704
1724
  "limit": {
1705
1725
  "context": 131072,
1706
1726
  "output": 8192
1707
1727
  },
1708
1728
  "cost": {
1709
- "input": 0.8,
1710
- "output": 2.4
1729
+ "input": 0.18,
1730
+ "output": 0.7,
1731
+ "reasoning": 2.1
1711
1732
  }
1712
1733
  },
1713
- "qwen3.5-122b-a10b": {
1714
- "id": "qwen3.5-122b-a10b",
1715
- "name": "Qwen3.5 122B-A10B",
1716
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1734
+ "qwen3-235b-a22b": {
1735
+ "id": "qwen3-235b-a22b",
1736
+ "name": "Qwen3 235B-A22B",
1737
+ "description": "Large open Qwen MoE for multilingual reasoning, coding, and tool use",
1717
1738
  "family": "qwen",
1718
- "attachment": true,
1739
+ "attachment": false,
1719
1740
  "reasoning": true,
1720
1741
  "reasoning_options": [
1721
1742
  {
@@ -1726,16 +1747,13 @@
1726
1747
  }
1727
1748
  ],
1728
1749
  "tool_call": true,
1729
- "structured_output": true,
1730
1750
  "temperature": true,
1731
- "release_date": "2026-02-23",
1732
- "last_updated": "2026-02-23",
1751
+ "knowledge": "2025-04",
1752
+ "release_date": "2025-04",
1753
+ "last_updated": "2025-04",
1733
1754
  "modalities": {
1734
1755
  "input": [
1735
- "text",
1736
- "image",
1737
- "video",
1738
- "audio"
1756
+ "text"
1739
1757
  ],
1740
1758
  "output": [
1741
1759
  "text"
@@ -1743,47 +1761,78 @@
1743
1761
  },
1744
1762
  "open_weights": true,
1745
1763
  "limit": {
1746
- "context": 262144,
1747
- "output": 65536
1764
+ "context": 131072,
1765
+ "output": 16384
1748
1766
  },
1749
1767
  "cost": {
1750
- "input": 0.4,
1751
- "output": 3.2
1768
+ "input": 0.7,
1769
+ "output": 2.8,
1770
+ "reasoning": 8.4
1752
1771
  }
1753
1772
  },
1754
- "qwen3-coder-plus": {
1755
- "id": "qwen3-coder-plus",
1756
- "name": "Qwen3 Coder Plus",
1757
- "description": "Hosted Qwen coder for software agents, repo edits, and long-context code",
1773
+ "qwen3.7-plus": {
1774
+ "id": "qwen3.7-plus",
1775
+ "name": "Qwen3.7 Plus",
1776
+ "description": "Multimodal Qwen workhorse for long-context agents, visual inputs, and coding",
1758
1777
  "family": "qwen",
1759
1778
  "attachment": false,
1760
- "reasoning": false,
1779
+ "reasoning": true,
1780
+ "reasoning_options": [
1781
+ {
1782
+ "type": "toggle"
1783
+ },
1784
+ {
1785
+ "type": "budget_tokens"
1786
+ }
1787
+ ],
1761
1788
  "tool_call": true,
1762
1789
  "temperature": true,
1763
1790
  "knowledge": "2025-04",
1764
- "release_date": "2025-07-23",
1765
- "last_updated": "2025-07-23",
1791
+ "release_date": "2026-06-02",
1792
+ "last_updated": "2026-06-04",
1766
1793
  "modalities": {
1767
1794
  "input": [
1768
- "text"
1795
+ "text",
1796
+ "image",
1797
+ "video"
1769
1798
  ],
1770
1799
  "output": [
1771
1800
  "text"
1772
1801
  ]
1773
1802
  },
1774
- "open_weights": true,
1803
+ "open_weights": false,
1775
1804
  "limit": {
1776
- "context": 1048576,
1805
+ "context": 1000000,
1777
1806
  "output": 65536
1778
1807
  },
1779
1808
  "cost": {
1780
- "input": 1,
1781
- "output": 5
1809
+ "input": 0.5,
1810
+ "output": 3,
1811
+ "cache_read": 0.05,
1812
+ "cache_write": 0.625,
1813
+ "tiers": [
1814
+ {
1815
+ "input": 2,
1816
+ "output": 6,
1817
+ "cache_read": 0.2,
1818
+ "cache_write": 2.5,
1819
+ "tier": {
1820
+ "type": "context",
1821
+ "size": 256000
1822
+ }
1823
+ }
1824
+ ],
1825
+ "context_over_200k": {
1826
+ "input": 2,
1827
+ "output": 6,
1828
+ "cache_read": 0.2,
1829
+ "cache_write": 2.5
1830
+ }
1782
1831
  }
1783
1832
  },
1784
- "qwen-omni-turbo-realtime": {
1785
- "id": "qwen-omni-turbo-realtime",
1786
- "name": "Qwen-Omni Turbo Realtime",
1833
+ "qwen-omni-turbo": {
1834
+ "id": "qwen-omni-turbo",
1835
+ "name": "Qwen-Omni Turbo",
1787
1836
  "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1788
1837
  "family": "qwen",
1789
1838
  "attachment": false,
@@ -1791,13 +1840,14 @@
1791
1840
  "tool_call": true,
1792
1841
  "temperature": true,
1793
1842
  "knowledge": "2024-04",
1794
- "release_date": "2025-05-08",
1795
- "last_updated": "2025-05-08",
1843
+ "release_date": "2025-01-19",
1844
+ "last_updated": "2025-03-26",
1796
1845
  "modalities": {
1797
1846
  "input": [
1798
1847
  "text",
1799
1848
  "image",
1800
- "audio"
1849
+ "audio",
1850
+ "video"
1801
1851
  ],
1802
1852
  "output": [
1803
1853
  "text",
@@ -1810,40 +1860,24 @@
1810
1860
  "output": 2048
1811
1861
  },
1812
1862
  "cost": {
1813
- "input": 0.27,
1814
- "output": 1.07,
1863
+ "input": 0.07,
1864
+ "output": 0.27,
1815
1865
  "input_audio": 4.44,
1816
1866
  "output_audio": 8.89
1817
1867
  }
1818
1868
  },
1819
- "deepseek-v4-flash-0731": {
1820
- "id": "deepseek-v4-flash-0731",
1821
- "name": "DeepSeek V4 Flash 0731",
1822
- "description": "Official DeepSeek V4 Flash release with enhanced agentic capabilities and integrated DSpark speculative decoding",
1823
- "family": "deepseek-flash",
1869
+ "qwen3-coder-480b-a35b-instruct": {
1870
+ "id": "qwen3-coder-480b-a35b-instruct",
1871
+ "name": "Qwen3-Coder 480B-A35B Instruct",
1872
+ "description": "Open Qwen coding heavyweight for repository reasoning and agentic engineering",
1873
+ "family": "qwen",
1824
1874
  "attachment": false,
1825
- "reasoning": true,
1826
- "reasoning_options": [
1827
- {
1828
- "type": "toggle"
1829
- },
1830
- {
1831
- "type": "effort",
1832
- "values": [
1833
- "high",
1834
- "max"
1835
- ]
1836
- }
1837
- ],
1875
+ "reasoning": false,
1838
1876
  "tool_call": true,
1839
- "interleaved": {
1840
- "field": "reasoning_content"
1841
- },
1842
- "structured_output": true,
1843
1877
  "temperature": true,
1844
- "knowledge": "2025-05",
1845
- "release_date": "2026-07-31",
1846
- "last_updated": "2026-07-31",
1878
+ "knowledge": "2025-04",
1879
+ "release_date": "2025-04",
1880
+ "last_updated": "2025-04",
1847
1881
  "modalities": {
1848
1882
  "input": [
1849
1883
  "text"
@@ -1854,43 +1888,30 @@
1854
1888
  },
1855
1889
  "open_weights": true,
1856
1890
  "limit": {
1857
- "context": 1000000,
1858
- "output": 384000
1891
+ "context": 262144,
1892
+ "output": 65536
1859
1893
  },
1860
1894
  "cost": {
1861
- "input": 0.2,
1862
- "output": 0.4,
1863
- "cache_read": 0.04
1864
- }
1865
- },
1866
- "qwen3-asr-flash": {
1867
- "id": "qwen3-asr-flash",
1868
- "name": "Qwen3-ASR Flash",
1869
- "description": "Speech transcription model for accurate audio-to-text and captioning workflows",
1870
- "family": "qwen",
1871
- "attachment": false,
1872
- "reasoning": false,
1873
- "tool_call": false,
1874
- "temperature": false,
1875
- "knowledge": "2024-04",
1876
- "release_date": "2025-09-08",
1877
- "last_updated": "2025-09-08",
1878
- "modalities": {
1879
- "input": [
1880
- "audio"
1881
- ],
1882
- "output": [
1883
- "text"
1895
+ "input": 1.5,
1896
+ "output": 7.5,
1897
+ "tiers": [
1898
+ {
1899
+ "input": 2.7,
1900
+ "output": 13.5,
1901
+ "tier": {
1902
+ "type": "context",
1903
+ "size": 32000
1904
+ }
1905
+ },
1906
+ {
1907
+ "input": 4.5,
1908
+ "output": 22.5,
1909
+ "tier": {
1910
+ "type": "context",
1911
+ "size": 128000
1912
+ }
1913
+ }
1884
1914
  ]
1885
- },
1886
- "open_weights": false,
1887
- "limit": {
1888
- "context": 53248,
1889
- "output": 4096
1890
- },
1891
- "cost": {
1892
- "input": 0.035,
1893
- "output": 0.035
1894
1915
  }
1895
1916
  },
1896
1917
  "qwen2-5-32b-instruct": {
@@ -1923,18 +1944,23 @@
1923
1944
  "output": 2.8
1924
1945
  }
1925
1946
  },
1926
- "qwen-vl-plus": {
1927
- "id": "qwen-vl-plus",
1928
- "name": "Qwen-VL Plus",
1947
+ "qwen3-vl-235b-a22b": {
1948
+ "id": "qwen3-vl-235b-a22b",
1949
+ "name": "Qwen3-VL 235B-A22B",
1929
1950
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1930
1951
  "family": "qwen",
1931
1952
  "attachment": false,
1932
- "reasoning": false,
1953
+ "reasoning": true,
1954
+ "reasoning_options": [
1955
+ {
1956
+ "type": "budget_tokens"
1957
+ }
1958
+ ],
1933
1959
  "tool_call": true,
1934
1960
  "temperature": true,
1935
- "knowledge": "2024-04",
1936
- "release_date": "2024-01-25",
1937
- "last_updated": "2025-08-15",
1961
+ "knowledge": "2025-04",
1962
+ "release_date": "2025-04",
1963
+ "last_updated": "2025-04",
1938
1964
  "modalities": {
1939
1965
  "input": [
1940
1966
  "text",
@@ -1944,22 +1970,23 @@
1944
1970
  "text"
1945
1971
  ]
1946
1972
  },
1947
- "open_weights": false,
1973
+ "open_weights": true,
1948
1974
  "limit": {
1949
1975
  "context": 131072,
1950
- "output": 8192
1976
+ "output": 32768
1951
1977
  },
1952
1978
  "cost": {
1953
- "input": 0.21,
1954
- "output": 0.63
1979
+ "input": 0.7,
1980
+ "output": 2.8,
1981
+ "reasoning": 8.4
1955
1982
  }
1956
1983
  },
1957
- "qwen3.6-flash": {
1958
- "id": "qwen3.6-flash",
1959
- "name": "Qwen3.6 Flash",
1984
+ "qwen3.5-plus": {
1985
+ "id": "qwen3.5-plus",
1986
+ "name": "Qwen3.5 Plus",
1960
1987
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1961
- "family": "qwen3.6",
1962
- "attachment": true,
1988
+ "family": "qwen",
1989
+ "attachment": false,
1963
1990
  "reasoning": true,
1964
1991
  "reasoning_options": [
1965
1992
  {
@@ -1970,10 +1997,10 @@
1970
1997
  }
1971
1998
  ],
1972
1999
  "tool_call": true,
1973
- "structured_output": true,
1974
2000
  "temperature": true,
1975
- "release_date": "2026-04-27",
1976
- "last_updated": "2026-04-27",
2001
+ "knowledge": "2025-04",
2002
+ "release_date": "2026-02-16",
2003
+ "last_updated": "2026-02-16",
1977
2004
  "modalities": {
1978
2005
  "input": [
1979
2006
  "text",
@@ -1990,9 +2017,71 @@
1990
2017
  "output": 65536
1991
2018
  },
1992
2019
  "cost": {
1993
- "input": 0.1875,
1994
- "output": 1.125,
1995
- "cache_write": 0.234375
2020
+ "input": 0.4,
2021
+ "output": 2.4,
2022
+ "reasoning": 2.4
2023
+ }
2024
+ },
2025
+ "qwen-mt-plus": {
2026
+ "id": "qwen-mt-plus",
2027
+ "name": "Qwen-MT Plus",
2028
+ "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
2029
+ "family": "qwen",
2030
+ "attachment": false,
2031
+ "reasoning": false,
2032
+ "tool_call": false,
2033
+ "temperature": true,
2034
+ "knowledge": "2024-04",
2035
+ "release_date": "2025-01",
2036
+ "last_updated": "2025-01",
2037
+ "modalities": {
2038
+ "input": [
2039
+ "text"
2040
+ ],
2041
+ "output": [
2042
+ "text"
2043
+ ]
2044
+ },
2045
+ "open_weights": false,
2046
+ "limit": {
2047
+ "context": 16384,
2048
+ "output": 8192
2049
+ },
2050
+ "cost": {
2051
+ "input": 2.46,
2052
+ "output": 7.37
2053
+ }
2054
+ },
2055
+ "qvq-max": {
2056
+ "id": "qvq-max",
2057
+ "name": "QVQ Max",
2058
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
2059
+ "family": "qvq",
2060
+ "attachment": false,
2061
+ "reasoning": true,
2062
+ "reasoning_options": [],
2063
+ "tool_call": true,
2064
+ "temperature": true,
2065
+ "knowledge": "2024-04",
2066
+ "release_date": "2025-03-25",
2067
+ "last_updated": "2025-03-25",
2068
+ "modalities": {
2069
+ "input": [
2070
+ "text",
2071
+ "image"
2072
+ ],
2073
+ "output": [
2074
+ "text"
2075
+ ]
2076
+ },
2077
+ "open_weights": false,
2078
+ "limit": {
2079
+ "context": 131072,
2080
+ "output": 8192
2081
+ },
2082
+ "cost": {
2083
+ "input": 1.2,
2084
+ "output": 4.8
1996
2085
  }
1997
2086
  }
1998
2087
  }