llm.rb 15.5.0 → 16.0.0.pre.2

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 (37) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +88 -1
  3. data/README.md +21 -30
  4. data/data/alibaba.json +1123 -949
  5. data/data/anthropic.json +296 -242
  6. data/data/bedrock.json +4178 -3807
  7. data/data/deepinfra.json +1529 -1358
  8. data/data/deepseek.json +15 -11
  9. data/data/google.json +787 -763
  10. data/data/mistral.json +476 -496
  11. data/data/moonshot.json +53 -50
  12. data/data/openai.json +1294 -1165
  13. data/data/openrouter.json +8722 -8546
  14. data/data/xai.json +207 -202
  15. data/data/zai.json +271 -264
  16. data/docs/deepdive/features/builtin_tools.md +3 -1
  17. data/docs/deepdive/features/console.md +7 -0
  18. data/docs/deepdive/fundamentals/schema.md +46 -0
  19. data/docs/deepdive/fundamentals/tools.md +68 -0
  20. data/docs/deepdive/protocols/a2a.md +91 -0
  21. data/docs/deepdive/protocols/mcp.md +50 -0
  22. data/docs/deepdive/reference/model_registry.md +1 -1
  23. data/lib/llm/console/input.rb +15 -5
  24. data/lib/llm/function/async/reactor.rb +10 -0
  25. data/lib/llm/function/async/task.rb +63 -11
  26. data/lib/llm/function/fork/job.rb +31 -8
  27. data/lib/llm/function/ractor/job.rb +59 -20
  28. data/lib/llm/function/ractor/mailbox.rb +39 -3
  29. data/lib/llm/function/ractor/task.rb +26 -9
  30. data/lib/llm/function/window.rb +77 -0
  31. data/lib/llm/function.rb +2 -1
  32. data/lib/llm/providers/deepseek/request_adapter/completion.rb +43 -2
  33. data/lib/llm/tool/param.rb +8 -8
  34. data/lib/llm/tools/exec.rb +2 -0
  35. data/lib/llm/version.rb +1 -1
  36. data/llm.gemspec +4 -4
  37. metadata +7 -6
data/data/alibaba.json CHANGED
@@ -8,10 +8,10 @@
8
8
  "name": "Alibaba",
9
9
  "doc": "https://www.alibabacloud.com/help/en/model-studio/models",
10
10
  "models": {
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",
11
+ "qwen-flash": {
12
+ "id": "qwen-flash",
13
+ "name": "Qwen Flash",
14
+ "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
15
15
  "family": "qwen",
16
16
  "attachment": false,
17
17
  "reasoning": true,
@@ -25,8 +25,9 @@
25
25
  ],
26
26
  "tool_call": true,
27
27
  "temperature": true,
28
- "release_date": "2026-05-21",
29
- "last_updated": "2026-05-21",
28
+ "knowledge": "2024-04",
29
+ "release_date": "2025-07-28",
30
+ "last_updated": "2025-07-28",
30
31
  "modalities": {
31
32
  "input": [
32
33
  "text"
@@ -38,62 +39,27 @@
38
39
  "open_weights": false,
39
40
  "limit": {
40
41
  "context": 1000000,
41
- "output": 131072
42
+ "output": 32768
42
43
  },
43
44
  "cost": {
44
- "input": 2.5,
45
- "output": 7.5,
46
- "cache_read": 0.5,
47
- "cache_write": 3.125
45
+ "input": 0.05,
46
+ "output": 0.4
48
47
  }
49
48
  },
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",
49
+ "qwen3.5-flash": {
50
+ "id": "qwen3.5-flash",
51
+ "name": "Qwen3.5 Flash",
52
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
54
53
  "family": "qwen",
55
- "attachment": false,
56
- "reasoning": false,
57
- "tool_call": true,
58
- "temperature": true,
59
- "knowledge": "2024-04",
60
- "release_date": "2024-09",
61
- "last_updated": "2024-09",
62
- "modalities": {
63
- "input": [
64
- "text"
65
- ],
66
- "output": [
67
- "text"
68
- ]
69
- },
70
- "open_weights": true,
71
- "limit": {
72
- "context": 131072,
73
- "output": 8192
74
- },
75
- "cost": {
76
- "input": 1.4,
77
- "output": 5.6
78
- }
79
- },
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",
85
- "attachment": false,
54
+ "attachment": true,
86
55
  "reasoning": true,
87
56
  "reasoning_options": [
88
57
  {
89
58
  "type": "toggle"
90
59
  },
91
60
  {
92
- "type": "effort",
93
- "values": [
94
- "high",
95
- "max"
96
- ]
61
+ "type": "budget_tokens",
62
+ "max": 81920
97
63
  }
98
64
  ],
99
65
  "tool_call": true,
@@ -102,75 +68,50 @@
102
68
  },
103
69
  "structured_output": true,
104
70
  "temperature": true,
105
- "knowledge": "2025-05",
106
- "release_date": "2026-07-31",
107
- "last_updated": "2026-07-31",
71
+ "release_date": "2026-02-23",
72
+ "last_updated": "2026-02-23",
108
73
  "modalities": {
109
74
  "input": [
110
- "text"
75
+ "text",
76
+ "image",
77
+ "video"
111
78
  ],
112
79
  "output": [
113
80
  "text"
114
81
  ]
115
82
  },
116
- "open_weights": true,
83
+ "open_weights": false,
117
84
  "limit": {
118
85
  "context": 1000000,
119
- "output": 384000
86
+ "output": 65536
120
87
  },
121
88
  "cost": {
122
- "input": 0.2,
89
+ "input": 0.1,
123
90
  "output": 0.4,
124
- "cache_read": 0.04
125
- }
126
- },
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",
131
- "family": "qwen",
132
- "attachment": false,
133
- "reasoning": false,
134
- "tool_call": true,
135
- "temperature": true,
136
- "knowledge": "2025-04",
137
- "release_date": "2025-07-23",
138
- "last_updated": "2025-07-23",
139
- "modalities": {
140
- "input": [
141
- "text"
142
- ],
143
- "output": [
144
- "text"
145
- ]
146
- },
147
- "open_weights": true,
148
- "limit": {
149
- "context": 1048576,
150
- "output": 65536
91
+ "cache_read": 0.01,
92
+ "cache_write": 0.125
151
93
  },
152
- "cost": {
153
- "input": 1,
154
- "output": 5
155
- }
94
+ "canonical_model_id": "alibaba/qwen3.5-flash"
156
95
  },
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",
96
+ "qwen3.7-max": {
97
+ "id": "qwen3.7-max",
98
+ "name": "Qwen3.7 Max",
99
+ "description": "Qwen frontier model tuned for agent frameworks, coding assistants, and long tasks",
161
100
  "family": "qwen",
162
101
  "attachment": false,
163
102
  "reasoning": true,
164
103
  "reasoning_options": [
104
+ {
105
+ "type": "toggle"
106
+ },
165
107
  {
166
108
  "type": "budget_tokens"
167
109
  }
168
110
  ],
169
111
  "tool_call": true,
170
112
  "temperature": true,
171
- "knowledge": "2025-04",
172
- "release_date": "2025-09",
173
- "last_updated": "2025-09",
113
+ "release_date": "2026-05-21",
114
+ "last_updated": "2026-05-21",
174
115
  "modalities": {
175
116
  "input": [
176
117
  "text"
@@ -179,63 +120,61 @@
179
120
  "text"
180
121
  ]
181
122
  },
182
- "open_weights": true,
123
+ "open_weights": false,
183
124
  "limit": {
184
- "context": 131072,
185
- "output": 32768
125
+ "context": 1000000,
126
+ "output": 131072
186
127
  },
187
128
  "cost": {
188
- "input": 0.5,
189
- "output": 6
129
+ "input": 2.5,
130
+ "output": 7.5,
131
+ "cache_read": 0.5,
132
+ "cache_write": 3.125
190
133
  }
191
134
  },
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",
135
+ "qwen2-5-32b-instruct": {
136
+ "id": "qwen2-5-32b-instruct",
137
+ "name": "Qwen2.5 32B Instruct",
138
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
196
139
  "family": "qwen",
197
140
  "attachment": false,
198
141
  "reasoning": false,
199
142
  "tool_call": true,
200
143
  "temperature": true,
201
144
  "knowledge": "2024-04",
202
- "release_date": "2024-12",
203
- "last_updated": "2024-12",
145
+ "release_date": "2024-09",
146
+ "last_updated": "2024-09",
204
147
  "modalities": {
205
148
  "input": [
206
- "text",
207
- "image",
208
- "audio",
209
- "video"
149
+ "text"
210
150
  ],
211
151
  "output": [
212
- "text",
213
- "audio"
152
+ "text"
214
153
  ]
215
154
  },
216
155
  "open_weights": true,
217
156
  "limit": {
218
- "context": 32768,
219
- "output": 2048
157
+ "context": 131072,
158
+ "output": 8192
220
159
  },
221
160
  "cost": {
222
- "input": 0.1,
223
- "output": 0.4,
224
- "input_audio": 6.76
161
+ "input": 0.7,
162
+ "output": 2.8
225
163
  }
226
164
  },
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",
165
+ "qwq-plus": {
166
+ "id": "qwq-plus",
167
+ "name": "QwQ Plus",
168
+ "description": "Qwen reasoning model for deliberate problem solving, math, and coding",
231
169
  "family": "qwen",
232
170
  "attachment": false,
233
- "reasoning": false,
234
- "tool_call": false,
171
+ "reasoning": true,
172
+ "reasoning_options": [],
173
+ "tool_call": true,
235
174
  "temperature": true,
236
175
  "knowledge": "2024-04",
237
- "release_date": "2025-01",
238
- "last_updated": "2025-01",
176
+ "release_date": "2025-03-05",
177
+ "last_updated": "2025-03-05",
239
178
  "modalities": {
240
179
  "input": [
241
180
  "text"
@@ -246,17 +185,17 @@
246
185
  },
247
186
  "open_weights": false,
248
187
  "limit": {
249
- "context": 16384,
188
+ "context": 131072,
250
189
  "output": 8192
251
190
  },
252
191
  "cost": {
253
- "input": 0.16,
254
- "output": 0.49
192
+ "input": 0.8,
193
+ "output": 2.4
255
194
  }
256
195
  },
257
- "qwen-vl-max": {
258
- "id": "qwen-vl-max",
259
- "name": "Qwen-VL Max",
196
+ "qwen2-5-vl-72b-instruct": {
197
+ "id": "qwen2-5-vl-72b-instruct",
198
+ "name": "Qwen2.5-VL 72B Instruct",
260
199
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
261
200
  "family": "qwen",
262
201
  "attachment": false,
@@ -264,8 +203,8 @@
264
203
  "tool_call": true,
265
204
  "temperature": true,
266
205
  "knowledge": "2024-04",
267
- "release_date": "2024-04-08",
268
- "last_updated": "2025-08-13",
206
+ "release_date": "2024-09",
207
+ "last_updated": "2024-09",
269
208
  "modalities": {
270
209
  "input": [
271
210
  "text",
@@ -275,81 +214,102 @@
275
214
  "text"
276
215
  ]
277
216
  },
278
- "open_weights": false,
217
+ "open_weights": true,
279
218
  "limit": {
280
219
  "context": 131072,
281
220
  "output": 8192
282
221
  },
283
222
  "cost": {
284
- "input": 0.8,
285
- "output": 3.2
223
+ "input": 2.8,
224
+ "output": 8.4
286
225
  }
287
226
  },
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",
227
+ "qwen3-vl-plus": {
228
+ "id": "qwen3-vl-plus",
229
+ "name": "Qwen3-VL Plus",
230
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
292
231
  "family": "qwen",
293
232
  "attachment": false,
294
- "reasoning": false,
233
+ "reasoning": true,
234
+ "reasoning_options": [
235
+ {
236
+ "type": "toggle"
237
+ },
238
+ {
239
+ "type": "budget_tokens"
240
+ }
241
+ ],
295
242
  "tool_call": true,
296
243
  "temperature": true,
297
244
  "knowledge": "2025-04",
298
- "release_date": "2025-09",
299
- "last_updated": "2025-09",
245
+ "release_date": "2025-09-23",
246
+ "last_updated": "2025-09-23",
300
247
  "modalities": {
301
248
  "input": [
302
- "text"
249
+ "text",
250
+ "image"
303
251
  ],
304
252
  "output": [
305
253
  "text"
306
254
  ]
307
255
  },
308
- "open_weights": true,
256
+ "open_weights": false,
309
257
  "limit": {
310
- "context": 131072,
258
+ "context": 262144,
311
259
  "output": 32768
312
260
  },
313
261
  "cost": {
314
- "input": 0.5,
315
- "output": 2
262
+ "input": 0.2,
263
+ "output": 1.6,
264
+ "reasoning": 4.8
316
265
  }
317
266
  },
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",
267
+ "qwen3.5-27b": {
268
+ "id": "qwen3.5-27b",
269
+ "name": "Qwen3.5 27B",
270
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
322
271
  "family": "qwen",
323
- "attachment": false,
324
- "reasoning": false,
325
- "tool_call": true,
326
- "temperature": true,
327
- "knowledge": "2025-04",
328
- "release_date": "2025-07-28",
329
- "last_updated": "2025-07-28",
330
- "modalities": {
272
+ "attachment": true,
273
+ "reasoning": true,
274
+ "reasoning_options": [
275
+ {
276
+ "type": "toggle"
277
+ },
278
+ {
279
+ "type": "budget_tokens"
280
+ }
281
+ ],
282
+ "tool_call": true,
283
+ "structured_output": true,
284
+ "temperature": true,
285
+ "release_date": "2026-02-23",
286
+ "last_updated": "2026-02-23",
287
+ "modalities": {
331
288
  "input": [
332
- "text"
289
+ "text",
290
+ "image",
291
+ "video",
292
+ "audio"
333
293
  ],
334
294
  "output": [
335
295
  "text"
336
296
  ]
337
297
  },
338
- "open_weights": false,
298
+ "open_weights": true,
339
299
  "limit": {
340
- "context": 1000000,
300
+ "context": 262144,
341
301
  "output": 65536
342
302
  },
343
303
  "cost": {
344
304
  "input": 0.3,
345
- "output": 1.5
305
+ "output": 2.4
346
306
  }
347
307
  },
348
- "qwen3-14b": {
349
- "id": "qwen3-14b",
350
- "name": "Qwen3 14B",
351
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
352
- "family": "qwen",
308
+ "deepseek-v4-flash-0731": {
309
+ "id": "deepseek-v4-flash-0731",
310
+ "name": "DeepSeek V4 Flash 0731",
311
+ "description": "Official DeepSeek V4 Flash release with enhanced agentic capabilities and integrated DSpark speculative decoding",
312
+ "family": "deepseek-flash",
353
313
  "attachment": false,
354
314
  "reasoning": true,
355
315
  "reasoning_options": [
@@ -357,14 +317,22 @@
357
317
  "type": "toggle"
358
318
  },
359
319
  {
360
- "type": "budget_tokens"
320
+ "type": "effort",
321
+ "values": [
322
+ "high",
323
+ "max"
324
+ ]
361
325
  }
362
326
  ],
363
327
  "tool_call": true,
328
+ "interleaved": {
329
+ "field": "reasoning_content"
330
+ },
331
+ "structured_output": true,
364
332
  "temperature": true,
365
- "knowledge": "2025-04",
366
- "release_date": "2025-04",
367
- "last_updated": "2025-04",
333
+ "knowledge": "2025-05",
334
+ "release_date": "2026-07-31",
335
+ "last_updated": "2026-07-31",
368
336
  "modalities": {
369
337
  "input": [
370
338
  "text"
@@ -375,14 +343,15 @@
375
343
  },
376
344
  "open_weights": true,
377
345
  "limit": {
378
- "context": 131072,
379
- "output": 8192
346
+ "context": 1000000,
347
+ "output": 384000
380
348
  },
381
349
  "cost": {
382
- "input": 0.35,
383
- "output": 1.4,
384
- "reasoning": 4.2
385
- }
350
+ "input": 0.2,
351
+ "output": 0.4,
352
+ "cache_read": 0.04
353
+ },
354
+ "canonical_model_id": "deepseek/deepseek-v4-flash-0731"
386
355
  },
387
356
  "qwen-max": {
388
357
  "id": "qwen-max",
@@ -414,31 +383,80 @@
414
383
  "output": 6.4
415
384
  }
416
385
  },
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",
386
+ "qwen3-next-80b-a3b-thinking": {
387
+ "id": "qwen3-next-80b-a3b-thinking",
388
+ "name": "Qwen3-Next 80B-A3B (Thinking)",
389
+ "description": "Efficient Qwen thinking model for local reasoning, math, and coding agents",
421
390
  "family": "qwen",
422
391
  "attachment": false,
423
392
  "reasoning": true,
393
+ "reasoning_options": [
394
+ {
395
+ "type": "budget_tokens"
396
+ }
397
+ ],
398
+ "tool_call": true,
399
+ "temperature": true,
400
+ "knowledge": "2025-04",
401
+ "release_date": "2025-09",
402
+ "last_updated": "2025-09",
403
+ "modalities": {
404
+ "input": [
405
+ "text"
406
+ ],
407
+ "output": [
408
+ "text"
409
+ ]
410
+ },
411
+ "open_weights": true,
412
+ "limit": {
413
+ "context": 131072,
414
+ "output": 32768
415
+ },
416
+ "cost": {
417
+ "input": 0.5,
418
+ "output": 6
419
+ }
420
+ },
421
+ "qwen3.8-max": {
422
+ "id": "qwen3.8-max",
423
+ "name": "Qwen3.8 Max",
424
+ "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
425
+ "family": "qwen",
426
+ "attachment": true,
427
+ "reasoning": true,
424
428
  "reasoning_options": [
425
429
  {
426
430
  "type": "toggle"
427
431
  },
428
432
  {
429
- "type": "budget_tokens"
433
+ "type": "effort",
434
+ "values": [
435
+ "low",
436
+ "medium",
437
+ "xhigh"
438
+ ]
439
+ },
440
+ {
441
+ "type": "budget_tokens",
442
+ "min": 0,
443
+ "max": 262144
430
444
  }
431
445
  ],
432
446
  "tool_call": true,
447
+ "interleaved": {
448
+ "field": "reasoning_content"
449
+ },
450
+ "structured_output": true,
433
451
  "temperature": true,
434
- "knowledge": "2025-04",
435
- "release_date": "2026-04-02",
436
- "last_updated": "2026-04-02",
452
+ "release_date": "2026-08-03",
453
+ "last_updated": "2026-08-03",
437
454
  "modalities": {
438
455
  "input": [
439
456
  "text",
440
457
  "image",
441
- "video"
458
+ "video",
459
+ "pdf"
442
460
  ],
443
461
  "output": [
444
462
  "text"
@@ -447,45 +465,41 @@
447
465
  "open_weights": false,
448
466
  "limit": {
449
467
  "context": 1000000,
450
- "output": 65536
468
+ "output": 131072
451
469
  },
452
470
  "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
- }
471
+ "input": 2,
472
+ "output": 6,
473
+ "cache_read": 0.25,
474
+ "cache_write": 2.5
475
+ },
476
+ "canonical_model_id": "alibaba/qwen3.8-max"
476
477
  },
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,
478
+ "kimi-k3": {
479
+ "id": "kimi-k3",
480
+ "name": "Kimi K3",
481
+ "description": "Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work",
482
+ "family": "kimi-k3",
483
+ "attachment": true,
484
+ "reasoning": true,
485
+ "reasoning_options": [
486
+ {
487
+ "type": "effort",
488
+ "values": [
489
+ "low",
490
+ "high",
491
+ "max"
492
+ ]
493
+ }
494
+ ],
484
495
  "tool_call": true,
485
- "temperature": true,
486
- "knowledge": "2024-04",
487
- "release_date": "2024-01-25",
488
- "last_updated": "2025-08-15",
496
+ "interleaved": {
497
+ "field": "reasoning_content"
498
+ },
499
+ "structured_output": true,
500
+ "temperature": false,
501
+ "release_date": "2026-07-16",
502
+ "last_updated": "2026-07-16",
489
503
  "modalities": {
490
504
  "input": [
491
505
  "text",
@@ -495,57 +509,54 @@
495
509
  "text"
496
510
  ]
497
511
  },
498
- "open_weights": false,
512
+ "open_weights": true,
499
513
  "limit": {
500
- "context": 131072,
501
- "output": 8192
514
+ "context": 1048576,
515
+ "output": 1048576
502
516
  },
503
517
  "cost": {
504
- "input": 0.21,
505
- "output": 0.63
506
- }
518
+ "input": 3,
519
+ "output": 15,
520
+ "cache_read": 0.3
521
+ },
522
+ "canonical_model_id": "moonshotai/kimi-k3"
507
523
  },
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",
524
+ "qwen-plus-character-ja": {
525
+ "id": "qwen-plus-character-ja",
526
+ "name": "Qwen Plus Character (Japanese)",
527
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
512
528
  "family": "qwen",
513
529
  "attachment": false,
514
530
  "reasoning": false,
515
531
  "tool_call": true,
516
532
  "temperature": true,
517
533
  "knowledge": "2024-04",
518
- "release_date": "2025-05-08",
519
- "last_updated": "2025-05-08",
534
+ "release_date": "2024-01",
535
+ "last_updated": "2024-01",
520
536
  "modalities": {
521
537
  "input": [
522
- "text",
523
- "image",
524
- "audio"
538
+ "text"
525
539
  ],
526
540
  "output": [
527
- "text",
528
- "audio"
541
+ "text"
529
542
  ]
530
543
  },
531
544
  "open_weights": false,
532
545
  "limit": {
533
- "context": 32768,
534
- "output": 2048
546
+ "context": 8192,
547
+ "output": 512
535
548
  },
536
549
  "cost": {
537
- "input": 0.27,
538
- "output": 1.07,
539
- "input_audio": 4.44,
540
- "output_audio": 8.89
550
+ "input": 0.5,
551
+ "output": 1.4
541
552
  }
542
553
  },
543
- "qwen3.5-27b": {
544
- "id": "qwen3.5-27b",
545
- "name": "Qwen3.5 27B",
554
+ "qwen3.5-plus": {
555
+ "id": "qwen3.5-plus",
556
+ "name": "Qwen3.5 Plus",
546
557
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
547
558
  "family": "qwen",
548
- "attachment": true,
559
+ "attachment": false,
549
560
  "reasoning": true,
550
561
  "reasoning_options": [
551
562
  {
@@ -556,37 +567,103 @@
556
567
  }
557
568
  ],
558
569
  "tool_call": true,
559
- "structured_output": true,
560
570
  "temperature": true,
561
- "release_date": "2026-02-23",
562
- "last_updated": "2026-02-23",
571
+ "knowledge": "2025-04",
572
+ "release_date": "2026-02-16",
573
+ "last_updated": "2026-02-16",
563
574
  "modalities": {
564
575
  "input": [
565
576
  "text",
566
577
  "image",
567
- "video",
568
- "audio"
578
+ "video"
569
579
  ],
570
580
  "output": [
571
581
  "text"
572
582
  ]
573
583
  },
574
- "open_weights": true,
584
+ "open_weights": false,
575
585
  "limit": {
576
- "context": 262144,
586
+ "context": 1000000,
577
587
  "output": 65536
578
588
  },
579
589
  "cost": {
580
- "input": 0.3,
581
- "output": 2.4
590
+ "input": 0.4,
591
+ "output": 2.4,
592
+ "reasoning": 2.4
582
593
  }
583
594
  },
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",
595
+ "qwen3-livetranslate-flash-realtime": {
596
+ "id": "qwen3-livetranslate-flash-realtime",
597
+ "name": "Qwen3-LiveTranslate Flash Realtime",
598
+ "description": "Speech generation model for controllable voice, narration, and audio delivery",
588
599
  "family": "qwen",
589
- "attachment": true,
600
+ "attachment": false,
601
+ "reasoning": false,
602
+ "tool_call": false,
603
+ "temperature": true,
604
+ "knowledge": "2024-04",
605
+ "release_date": "2025-09-22",
606
+ "last_updated": "2025-09-22",
607
+ "modalities": {
608
+ "input": [
609
+ "text",
610
+ "image",
611
+ "audio",
612
+ "video"
613
+ ],
614
+ "output": [
615
+ "text",
616
+ "audio"
617
+ ]
618
+ },
619
+ "open_weights": false,
620
+ "limit": {
621
+ "context": 53248,
622
+ "output": 4096
623
+ },
624
+ "cost": {
625
+ "input": 10,
626
+ "output": 10,
627
+ "input_audio": 10,
628
+ "output_audio": 38
629
+ }
630
+ },
631
+ "qwen2-5-72b-instruct": {
632
+ "id": "qwen2-5-72b-instruct",
633
+ "name": "Qwen2.5 72B Instruct",
634
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
635
+ "family": "qwen",
636
+ "attachment": false,
637
+ "reasoning": false,
638
+ "tool_call": true,
639
+ "temperature": true,
640
+ "knowledge": "2024-04",
641
+ "release_date": "2024-09",
642
+ "last_updated": "2024-09",
643
+ "modalities": {
644
+ "input": [
645
+ "text"
646
+ ],
647
+ "output": [
648
+ "text"
649
+ ]
650
+ },
651
+ "open_weights": true,
652
+ "limit": {
653
+ "context": 131072,
654
+ "output": 8192
655
+ },
656
+ "cost": {
657
+ "input": 1.4,
658
+ "output": 5.6
659
+ }
660
+ },
661
+ "qwen3-32b": {
662
+ "id": "qwen3-32b",
663
+ "name": "Qwen3 32B",
664
+ "description": "Dense open Qwen model for self-hosted chat, reasoning, and coding",
665
+ "family": "qwen",
666
+ "attachment": false,
590
667
  "reasoning": true,
591
668
  "reasoning_options": [
592
669
  {
@@ -597,16 +674,13 @@
597
674
  }
598
675
  ],
599
676
  "tool_call": true,
600
- "structured_output": true,
601
677
  "temperature": true,
602
- "release_date": "2026-02-23",
603
- "last_updated": "2026-02-23",
678
+ "knowledge": "2025-04",
679
+ "release_date": "2025-04",
680
+ "last_updated": "2025-04",
604
681
  "modalities": {
605
682
  "input": [
606
- "text",
607
- "image",
608
- "video",
609
- "audio"
683
+ "text"
610
684
  ],
611
685
  "output": [
612
686
  "text"
@@ -614,18 +688,19 @@
614
688
  },
615
689
  "open_weights": true,
616
690
  "limit": {
617
- "context": 262144,
618
- "output": 65536
691
+ "context": 131072,
692
+ "output": 16384
619
693
  },
620
694
  "cost": {
621
- "input": 0.25,
622
- "output": 2
695
+ "input": 0.7,
696
+ "output": 2.8,
697
+ "reasoning": 8.4
623
698
  }
624
699
  },
625
- "qwen-flash": {
626
- "id": "qwen-flash",
627
- "name": "Qwen Flash",
628
- "description": "Efficient Qwen model for fast chat, extraction, and high-volume workloads",
700
+ "qwen-plus": {
701
+ "id": "qwen-plus",
702
+ "name": "Qwen Plus",
703
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
629
704
  "family": "qwen",
630
705
  "attachment": false,
631
706
  "reasoning": true,
@@ -640,8 +715,8 @@
640
715
  "tool_call": true,
641
716
  "temperature": true,
642
717
  "knowledge": "2024-04",
643
- "release_date": "2025-07-28",
644
- "last_updated": "2025-07-28",
718
+ "release_date": "2024-01-25",
719
+ "last_updated": "2025-09-11",
645
720
  "modalities": {
646
721
  "input": [
647
722
  "text"
@@ -656,29 +731,25 @@
656
731
  "output": 32768
657
732
  },
658
733
  "cost": {
659
- "input": 0.05,
660
- "output": 0.4
734
+ "input": 0.4,
735
+ "output": 1.2,
736
+ "reasoning": 4
661
737
  }
662
738
  },
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",
668
- "attachment": false,
739
+ "qwen3.7-flash": {
740
+ "id": "qwen3.7-flash",
741
+ "name": "Qwen3.7 Flash",
742
+ "description": "Lightweight multimodal Qwen model for high-throughput text, image, and video tasks",
743
+ "family": "qwen",
744
+ "attachment": true,
669
745
  "reasoning": true,
670
746
  "reasoning_options": [
671
747
  {
672
- "type": "effort",
673
- "values": [
674
- "none",
675
- "minimal",
676
- "low",
677
- "medium",
678
- "high",
679
- "xhigh",
680
- "max"
681
- ]
748
+ "type": "toggle"
749
+ },
750
+ {
751
+ "type": "budget_tokens",
752
+ "max": 262144
682
753
  }
683
754
  ],
684
755
  "tool_call": true,
@@ -687,26 +758,165 @@
687
758
  },
688
759
  "structured_output": true,
689
760
  "temperature": true,
690
- "release_date": "2026-06-13",
691
- "last_updated": "2026-06-13",
761
+ "release_date": "2026-07-15",
762
+ "last_updated": "2026-07-15",
692
763
  "modalities": {
693
764
  "input": [
694
- "text"
765
+ "text",
766
+ "image",
767
+ "video"
695
768
  ],
696
769
  "output": [
697
770
  "text"
698
771
  ]
699
772
  },
700
- "open_weights": true,
773
+ "open_weights": false,
701
774
  "limit": {
702
775
  "context": 1000000,
776
+ "input": 991808,
703
777
  "output": 131072
704
778
  },
705
779
  "cost": {
706
- "input": 1.4,
707
- "output": 4.4,
708
- "cache_read": 0.28,
709
- "cache_write": 0
780
+ "input": 0.03,
781
+ "output": 0.13,
782
+ "cache_read": 0.003,
783
+ "cache_write": 0.0375,
784
+ "tiers": [
785
+ {
786
+ "input": 0.1,
787
+ "output": 0.4,
788
+ "cache_read": 0.01,
789
+ "cache_write": 0.125,
790
+ "tier": {
791
+ "type": "context",
792
+ "size": 32000
793
+ }
794
+ },
795
+ {
796
+ "input": 0.2,
797
+ "output": 0.8,
798
+ "cache_read": 0.02,
799
+ "cache_write": 0.25,
800
+ "tier": {
801
+ "type": "context",
802
+ "size": 256000
803
+ }
804
+ }
805
+ ]
806
+ },
807
+ "canonical_model_id": "alibaba/qwen3.7-flash"
808
+ },
809
+ "qwen3-max": {
810
+ "id": "qwen3-max",
811
+ "name": "Qwen3 Max",
812
+ "description": "Flagship Qwen3 model for coding agents, complex reasoning, and tool use",
813
+ "family": "qwen",
814
+ "attachment": false,
815
+ "reasoning": false,
816
+ "tool_call": true,
817
+ "temperature": true,
818
+ "knowledge": "2025-04",
819
+ "release_date": "2025-09-23",
820
+ "last_updated": "2025-09-23",
821
+ "modalities": {
822
+ "input": [
823
+ "text"
824
+ ],
825
+ "output": [
826
+ "text"
827
+ ]
828
+ },
829
+ "open_weights": false,
830
+ "limit": {
831
+ "context": 262144,
832
+ "output": 65536
833
+ },
834
+ "cost": {
835
+ "input": 1.2,
836
+ "output": 6
837
+ }
838
+ },
839
+ "qwen-omni-turbo-realtime": {
840
+ "id": "qwen-omni-turbo-realtime",
841
+ "name": "Qwen-Omni Turbo Realtime",
842
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
843
+ "family": "qwen",
844
+ "attachment": false,
845
+ "reasoning": false,
846
+ "tool_call": true,
847
+ "temperature": true,
848
+ "knowledge": "2024-04",
849
+ "release_date": "2025-05-08",
850
+ "last_updated": "2025-05-08",
851
+ "modalities": {
852
+ "input": [
853
+ "text",
854
+ "image",
855
+ "audio"
856
+ ],
857
+ "output": [
858
+ "text",
859
+ "audio"
860
+ ]
861
+ },
862
+ "open_weights": false,
863
+ "limit": {
864
+ "context": 32768,
865
+ "output": 2048
866
+ },
867
+ "cost": {
868
+ "input": 0.27,
869
+ "output": 1.07,
870
+ "input_audio": 4.44,
871
+ "output_audio": 8.89
872
+ }
873
+ },
874
+ "qwen3-coder-480b-a35b-instruct": {
875
+ "id": "qwen3-coder-480b-a35b-instruct",
876
+ "name": "Qwen3-Coder 480B-A35B Instruct",
877
+ "description": "Open Qwen coding heavyweight for repository reasoning and agentic engineering",
878
+ "family": "qwen",
879
+ "attachment": false,
880
+ "reasoning": false,
881
+ "tool_call": true,
882
+ "temperature": true,
883
+ "knowledge": "2025-04",
884
+ "release_date": "2025-04",
885
+ "last_updated": "2025-04",
886
+ "modalities": {
887
+ "input": [
888
+ "text"
889
+ ],
890
+ "output": [
891
+ "text"
892
+ ]
893
+ },
894
+ "open_weights": true,
895
+ "limit": {
896
+ "context": 262144,
897
+ "output": 65536
898
+ },
899
+ "cost": {
900
+ "input": 1.5,
901
+ "output": 7.5,
902
+ "tiers": [
903
+ {
904
+ "input": 2.7,
905
+ "output": 13.5,
906
+ "tier": {
907
+ "type": "context",
908
+ "size": 32000
909
+ }
910
+ },
911
+ {
912
+ "input": 4.5,
913
+ "output": 22.5,
914
+ "tier": {
915
+ "type": "context",
916
+ "size": 128000
917
+ }
918
+ }
919
+ ]
710
920
  }
711
921
  },
712
922
  "qwen-turbo": {
@@ -748,91 +958,115 @@
748
958
  "reasoning": 0.5
749
959
  }
750
960
  },
751
- "qwen3-livetranslate-flash-realtime": {
752
- "id": "qwen3-livetranslate-flash-realtime",
753
- "name": "Qwen3-LiveTranslate Flash Realtime",
754
- "description": "Speech generation model for controllable voice, narration, and audio delivery",
961
+ "qwen3-coder-30b-a3b-instruct": {
962
+ "id": "qwen3-coder-30b-a3b-instruct",
963
+ "name": "Qwen3-Coder 30B-A3B Instruct",
964
+ "description": "Smaller Qwen coder for efficient local agents and repo-level fixes",
755
965
  "family": "qwen",
756
966
  "attachment": false,
757
967
  "reasoning": false,
758
- "tool_call": false,
968
+ "tool_call": true,
759
969
  "temperature": true,
760
- "knowledge": "2024-04",
761
- "release_date": "2025-09-22",
762
- "last_updated": "2025-09-22",
970
+ "knowledge": "2025-04",
971
+ "release_date": "2025-04",
972
+ "last_updated": "2025-04",
763
973
  "modalities": {
764
974
  "input": [
765
- "text",
766
- "image",
767
- "audio",
768
- "video"
975
+ "text"
769
976
  ],
770
977
  "output": [
771
- "text",
772
- "audio"
978
+ "text"
773
979
  ]
774
980
  },
775
- "open_weights": false,
981
+ "open_weights": true,
776
982
  "limit": {
777
- "context": 53248,
778
- "output": 4096
983
+ "context": 262144,
984
+ "output": 65536
779
985
  },
780
986
  "cost": {
781
- "input": 10,
782
- "output": 10,
783
- "input_audio": 10,
784
- "output_audio": 38
987
+ "input": 0.45,
988
+ "output": 2.25,
989
+ "tiers": [
990
+ {
991
+ "input": 0.75,
992
+ "output": 3.75,
993
+ "tier": {
994
+ "type": "context",
995
+ "size": 32000
996
+ }
997
+ },
998
+ {
999
+ "input": 1.2,
1000
+ "output": 6,
1001
+ "tier": {
1002
+ "type": "context",
1003
+ "size": 128000
1004
+ }
1005
+ }
1006
+ ]
785
1007
  }
786
1008
  },
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",
1009
+ "qwen3.8-omni-flash": {
1010
+ "id": "qwen3.8-omni-flash",
1011
+ "name": "Qwen3.8 Omni Flash",
1012
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
791
1013
  "family": "qwen",
792
- "attachment": false,
1014
+ "attachment": true,
793
1015
  "reasoning": true,
794
1016
  "reasoning_options": [
795
1017
  {
796
- "type": "budget_tokens"
1018
+ "type": "effort",
1019
+ "values": [
1020
+ "none",
1021
+ "low",
1022
+ "medium",
1023
+ "xhigh"
1024
+ ]
797
1025
  }
798
1026
  ],
799
1027
  "tool_call": true,
800
- "temperature": true,
801
- "knowledge": "2025-04",
802
- "release_date": "2025-04",
803
- "last_updated": "2025-04",
1028
+ "interleaved": {
1029
+ "field": "reasoning_content"
1030
+ },
1031
+ "structured_output": true,
1032
+ "release_date": "2026-09-17",
1033
+ "last_updated": "2026-09-17",
804
1034
  "modalities": {
805
1035
  "input": [
806
1036
  "text",
807
- "image"
1037
+ "image",
1038
+ "audio",
1039
+ "video"
808
1040
  ],
809
1041
  "output": [
810
1042
  "text"
811
1043
  ]
812
1044
  },
813
- "open_weights": true,
1045
+ "open_weights": false,
814
1046
  "limit": {
815
- "context": 131072,
816
- "output": 32768
1047
+ "context": 1000000,
1048
+ "output": 131072
817
1049
  },
818
1050
  "cost": {
819
- "input": 0.2,
820
- "output": 0.8,
821
- "reasoning": 2.4
822
- }
1051
+ "input": 0.15,
1052
+ "output": 0.47,
1053
+ "cache_read": 0.016
1054
+ },
1055
+ "canonical_model_id": "alibaba/qwen3.8-omni-flash"
823
1056
  },
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",
828
- "family": "qwen",
1057
+ "qvq-max": {
1058
+ "id": "qvq-max",
1059
+ "name": "QVQ Max",
1060
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1061
+ "family": "qvq",
829
1062
  "attachment": false,
830
- "reasoning": false,
831
- "tool_call": false,
1063
+ "reasoning": true,
1064
+ "reasoning_options": [],
1065
+ "tool_call": true,
832
1066
  "temperature": true,
833
1067
  "knowledge": "2024-04",
834
- "release_date": "2024-10-28",
835
- "last_updated": "2025-04-13",
1068
+ "release_date": "2025-03-25",
1069
+ "last_updated": "2025-03-25",
836
1070
  "modalities": {
837
1071
  "input": [
838
1072
  "text",
@@ -844,34 +1078,26 @@
844
1078
  },
845
1079
  "open_weights": false,
846
1080
  "limit": {
847
- "context": 34096,
848
- "output": 4096
1081
+ "context": 131072,
1082
+ "output": 8192
849
1083
  },
850
1084
  "cost": {
851
- "input": 0.72,
852
- "output": 0.72
1085
+ "input": 1.2,
1086
+ "output": 4.8
853
1087
  }
854
1088
  },
855
- "qwen3-32b": {
856
- "id": "qwen3-32b",
857
- "name": "Qwen3 32B",
858
- "description": "Dense open Qwen model for self-hosted chat, reasoning, and coding",
1089
+ "qwen2-5-7b-instruct": {
1090
+ "id": "qwen2-5-7b-instruct",
1091
+ "name": "Qwen2.5 7B Instruct",
1092
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
859
1093
  "family": "qwen",
860
1094
  "attachment": false,
861
- "reasoning": true,
862
- "reasoning_options": [
863
- {
864
- "type": "toggle"
865
- },
866
- {
867
- "type": "budget_tokens"
868
- }
869
- ],
1095
+ "reasoning": false,
870
1096
  "tool_call": true,
871
1097
  "temperature": true,
872
- "knowledge": "2025-04",
873
- "release_date": "2025-04",
874
- "last_updated": "2025-04",
1098
+ "knowledge": "2024-04",
1099
+ "release_date": "2024-09",
1100
+ "last_updated": "2024-09",
875
1101
  "modalities": {
876
1102
  "input": [
877
1103
  "text"
@@ -883,65 +1109,66 @@
883
1109
  "open_weights": true,
884
1110
  "limit": {
885
1111
  "context": 131072,
886
- "output": 16384
1112
+ "output": 8192
887
1113
  },
888
1114
  "cost": {
889
- "input": 0.7,
890
- "output": 2.8,
891
- "reasoning": 8.4
1115
+ "input": 0.175,
1116
+ "output": 0.7
892
1117
  }
893
1118
  },
894
- "qwq-plus": {
895
- "id": "qwq-plus",
896
- "name": "QwQ Plus",
897
- "description": "Qwen reasoning model for deliberate problem solving, math, and coding",
1119
+ "qwen3-omni-flash": {
1120
+ "id": "qwen3-omni-flash",
1121
+ "name": "Qwen3-Omni Flash",
1122
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
898
1123
  "family": "qwen",
899
1124
  "attachment": false,
900
1125
  "reasoning": true,
901
- "reasoning_options": [],
1126
+ "reasoning_options": [
1127
+ {
1128
+ "type": "toggle"
1129
+ }
1130
+ ],
902
1131
  "tool_call": true,
903
1132
  "temperature": true,
904
1133
  "knowledge": "2024-04",
905
- "release_date": "2025-03-05",
906
- "last_updated": "2025-03-05",
1134
+ "release_date": "2025-09-15",
1135
+ "last_updated": "2025-09-15",
907
1136
  "modalities": {
908
1137
  "input": [
909
- "text"
1138
+ "text",
1139
+ "image",
1140
+ "audio",
1141
+ "video"
910
1142
  ],
911
1143
  "output": [
912
- "text"
1144
+ "text",
1145
+ "audio"
913
1146
  ]
914
1147
  },
915
1148
  "open_weights": false,
916
1149
  "limit": {
917
- "context": 131072,
918
- "output": 8192
1150
+ "context": 65536,
1151
+ "output": 16384
919
1152
  },
920
1153
  "cost": {
921
- "input": 0.8,
922
- "output": 2.4
1154
+ "input": 0.43,
1155
+ "output": 1.66,
1156
+ "input_audio": 3.81,
1157
+ "output_audio": 15.11
923
1158
  }
924
1159
  },
925
- "qwen3-vl-plus": {
926
- "id": "qwen3-vl-plus",
927
- "name": "Qwen3-VL Plus",
1160
+ "qwen-vl-max": {
1161
+ "id": "qwen-vl-max",
1162
+ "name": "Qwen-VL Max",
928
1163
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
929
1164
  "family": "qwen",
930
1165
  "attachment": false,
931
- "reasoning": true,
932
- "reasoning_options": [
933
- {
934
- "type": "toggle"
935
- },
936
- {
937
- "type": "budget_tokens"
938
- }
939
- ],
1166
+ "reasoning": false,
940
1167
  "tool_call": true,
941
1168
  "temperature": true,
942
- "knowledge": "2025-04",
943
- "release_date": "2025-09-23",
944
- "last_updated": "2025-09-23",
1169
+ "knowledge": "2024-04",
1170
+ "release_date": "2024-04-08",
1171
+ "last_updated": "2025-08-13",
945
1172
  "modalities": {
946
1173
  "input": [
947
1174
  "text",
@@ -953,27 +1180,34 @@
953
1180
  },
954
1181
  "open_weights": false,
955
1182
  "limit": {
956
- "context": 262144,
957
- "output": 32768
1183
+ "context": 131072,
1184
+ "output": 8192
958
1185
  },
959
1186
  "cost": {
960
- "input": 0.2,
961
- "output": 1.6,
962
- "reasoning": 4.8
1187
+ "input": 0.8,
1188
+ "output": 3.2
963
1189
  }
964
1190
  },
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",
1191
+ "qwen3-235b-a22b": {
1192
+ "id": "qwen3-235b-a22b",
1193
+ "name": "Qwen3 235B-A22B",
1194
+ "description": "Large open Qwen MoE for multilingual reasoning, coding, and tool use",
969
1195
  "family": "qwen",
970
1196
  "attachment": false,
971
- "reasoning": false,
1197
+ "reasoning": true,
1198
+ "reasoning_options": [
1199
+ {
1200
+ "type": "toggle"
1201
+ },
1202
+ {
1203
+ "type": "budget_tokens"
1204
+ }
1205
+ ],
972
1206
  "tool_call": true,
973
1207
  "temperature": true,
974
- "knowledge": "2024-04",
975
- "release_date": "2024-09",
976
- "last_updated": "2024-09",
1208
+ "knowledge": "2025-04",
1209
+ "release_date": "2025-04",
1210
+ "last_updated": "2025-04",
977
1211
  "modalities": {
978
1212
  "input": [
979
1213
  "text"
@@ -985,20 +1219,26 @@
985
1219
  "open_weights": true,
986
1220
  "limit": {
987
1221
  "context": 131072,
988
- "output": 8192
1222
+ "output": 16384
989
1223
  },
990
1224
  "cost": {
991
- "input": 0.175,
992
- "output": 0.7
1225
+ "input": 0.7,
1226
+ "output": 2.8,
1227
+ "reasoning": 8.4
993
1228
  }
994
1229
  },
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",
1230
+ "qwen3-vl-30b-a3b": {
1231
+ "id": "qwen3-vl-30b-a3b",
1232
+ "name": "Qwen3-VL 30B-A3B",
1233
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
999
1234
  "family": "qwen",
1000
1235
  "attachment": false,
1001
- "reasoning": false,
1236
+ "reasoning": true,
1237
+ "reasoning_options": [
1238
+ {
1239
+ "type": "budget_tokens"
1240
+ }
1241
+ ],
1002
1242
  "tool_call": true,
1003
1243
  "temperature": true,
1004
1244
  "knowledge": "2025-04",
@@ -1006,7 +1246,8 @@
1006
1246
  "last_updated": "2025-04",
1007
1247
  "modalities": {
1008
1248
  "input": [
1009
- "text"
1249
+ "text",
1250
+ "image"
1010
1251
  ],
1011
1252
  "output": [
1012
1253
  "text"
@@ -1014,36 +1255,19 @@
1014
1255
  },
1015
1256
  "open_weights": true,
1016
1257
  "limit": {
1017
- "context": 262144,
1018
- "output": 65536
1258
+ "context": 131072,
1259
+ "output": 32768
1019
1260
  },
1020
1261
  "cost": {
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
- ]
1262
+ "input": 0.2,
1263
+ "output": 0.8,
1264
+ "reasoning": 2.4
1041
1265
  }
1042
1266
  },
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",
1267
+ "qwen3.5-122b-a10b": {
1268
+ "id": "qwen3.5-122b-a10b",
1269
+ "name": "Qwen3.5 122B-A10B",
1270
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1047
1271
  "family": "qwen",
1048
1272
  "attachment": true,
1049
1273
  "reasoning": true,
@@ -1058,8 +1282,8 @@
1058
1282
  "tool_call": true,
1059
1283
  "structured_output": true,
1060
1284
  "temperature": true,
1061
- "release_date": "2026-02-15",
1062
- "last_updated": "2026-02-15",
1285
+ "release_date": "2026-02-23",
1286
+ "last_updated": "2026-02-23",
1063
1287
  "modalities": {
1064
1288
  "input": [
1065
1289
  "text",
@@ -1077,14 +1301,14 @@
1077
1301
  "output": 65536
1078
1302
  },
1079
1303
  "cost": {
1080
- "input": 0.6,
1081
- "output": 3.6
1304
+ "input": 0.4,
1305
+ "output": 3.2
1082
1306
  }
1083
1307
  },
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",
1308
+ "qwen3.6-35b-a3b": {
1309
+ "id": "qwen3.6-35b-a3b",
1310
+ "name": "Qwen3.6 35B-A3B",
1311
+ "description": "Open multimodal Qwen MoE for local agents that need vision, audio, and code",
1088
1312
  "family": "qwen",
1089
1313
  "attachment": true,
1090
1314
  "reasoning": true,
@@ -1099,8 +1323,8 @@
1099
1323
  "tool_call": true,
1100
1324
  "structured_output": true,
1101
1325
  "temperature": true,
1102
- "release_date": "2026-04-22",
1103
- "last_updated": "2026-04-22",
1326
+ "release_date": "2026-04-17",
1327
+ "last_updated": "2026-04-17",
1104
1328
  "modalities": {
1105
1329
  "input": [
1106
1330
  "text",
@@ -1118,110 +1342,25 @@
1118
1342
  "output": 65536
1119
1343
  },
1120
1344
  "cost": {
1121
- "input": 0.6,
1122
- "output": 3.6
1345
+ "input": 0.248,
1346
+ "output": 1.485
1123
1347
  }
1124
1348
  },
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",
1349
+ "qwen3-next-80b-a3b-instruct": {
1350
+ "id": "qwen3-next-80b-a3b-instruct",
1351
+ "name": "Qwen3-Next 80B-A3B Instruct",
1352
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1129
1353
  "family": "qwen",
1130
1354
  "attachment": false,
1131
1355
  "reasoning": false,
1132
- "tool_call": false,
1133
- "temperature": false,
1134
- "knowledge": "2024-04",
1135
- "release_date": "2025-09-08",
1136
- "last_updated": "2025-09-08",
1137
- "modalities": {
1138
- "input": [
1139
- "audio"
1140
- ],
1141
- "output": [
1142
- "text"
1143
- ]
1144
- },
1145
- "open_weights": false,
1146
- "limit": {
1147
- "context": 53248,
1148
- "output": 4096
1149
- },
1150
- "cost": {
1151
- "input": 0.035,
1152
- "output": 0.035
1153
- }
1154
- },
1155
- "qwen3-omni-flash": {
1156
- "id": "qwen3-omni-flash",
1157
- "name": "Qwen3-Omni Flash",
1158
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1159
- "family": "qwen",
1160
- "attachment": false,
1161
- "reasoning": true,
1162
- "reasoning_options": [
1163
- {
1164
- "type": "toggle"
1165
- }
1166
- ],
1167
1356
  "tool_call": true,
1168
1357
  "temperature": true,
1169
- "knowledge": "2024-04",
1170
- "release_date": "2025-09-15",
1171
- "last_updated": "2025-09-15",
1172
- "modalities": {
1173
- "input": [
1174
- "text",
1175
- "image",
1176
- "audio",
1177
- "video"
1178
- ],
1179
- "output": [
1180
- "text",
1181
- "audio"
1182
- ]
1183
- },
1184
- "open_weights": false,
1185
- "limit": {
1186
- "context": 65536,
1187
- "output": 16384
1188
- },
1189
- "cost": {
1190
- "input": 0.43,
1191
- "output": 1.66,
1192
- "input_audio": 3.81,
1193
- "output_audio": 15.11
1194
- }
1195
- },
1196
- "kimi-k3": {
1197
- "id": "kimi-k3",
1198
- "name": "Kimi K3",
1199
- "description": "Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work",
1200
- "family": "kimi-k3",
1201
- "attachment": true,
1202
- "reasoning": true,
1203
- "reasoning_options": [
1204
- {
1205
- "type": "effort",
1206
- "values": [
1207
- "low",
1208
- "high",
1209
- "max"
1210
- ]
1211
- }
1212
- ],
1213
- "tool_call": true,
1214
- "interleaved": {
1215
- "field": "reasoning_content"
1216
- },
1217
- "structured_output": true,
1218
- "temperature": false,
1219
- "release_date": "2026-07-16",
1220
- "last_updated": "2026-07-16",
1358
+ "knowledge": "2025-04",
1359
+ "release_date": "2025-09",
1360
+ "last_updated": "2025-09",
1221
1361
  "modalities": {
1222
1362
  "input": [
1223
- "text",
1224
- "image"
1363
+ "text"
1225
1364
  ],
1226
1365
  "output": [
1227
1366
  "text"
@@ -1229,18 +1368,17 @@
1229
1368
  },
1230
1369
  "open_weights": true,
1231
1370
  "limit": {
1232
- "context": 1048576,
1233
- "output": 1048576
1371
+ "context": 131072,
1372
+ "output": 32768
1234
1373
  },
1235
1374
  "cost": {
1236
- "input": 3,
1237
- "output": 15,
1238
- "cache_read": 0.3
1375
+ "input": 0.5,
1376
+ "output": 2
1239
1377
  }
1240
1378
  },
1241
- "qwen2-5-vl-72b-instruct": {
1242
- "id": "qwen2-5-vl-72b-instruct",
1243
- "name": "Qwen2.5-VL 72B Instruct",
1379
+ "qwen-vl-plus": {
1380
+ "id": "qwen-vl-plus",
1381
+ "name": "Qwen-VL Plus",
1244
1382
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1245
1383
  "family": "qwen",
1246
1384
  "attachment": false,
@@ -1248,8 +1386,8 @@
1248
1386
  "tool_call": true,
1249
1387
  "temperature": true,
1250
1388
  "knowledge": "2024-04",
1251
- "release_date": "2024-09",
1252
- "last_updated": "2024-09",
1389
+ "release_date": "2024-01-25",
1390
+ "last_updated": "2025-08-15",
1253
1391
  "modalities": {
1254
1392
  "input": [
1255
1393
  "text",
@@ -1259,21 +1397,21 @@
1259
1397
  "text"
1260
1398
  ]
1261
1399
  },
1262
- "open_weights": true,
1400
+ "open_weights": false,
1263
1401
  "limit": {
1264
1402
  "context": 131072,
1265
1403
  "output": 8192
1266
1404
  },
1267
1405
  "cost": {
1268
- "input": 2.8,
1269
- "output": 8.4
1406
+ "input": 0.21,
1407
+ "output": 0.63
1270
1408
  }
1271
1409
  },
1272
- "qwen3.6-35b-a3b": {
1273
- "id": "qwen3.6-35b-a3b",
1274
- "name": "Qwen3.6 35B-A3B",
1275
- "description": "Open multimodal Qwen MoE for local agents that need vision, audio, and code",
1276
- "family": "qwen",
1410
+ "qwen3.6-flash": {
1411
+ "id": "qwen3.6-flash",
1412
+ "name": "Qwen3.6 Flash",
1413
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1414
+ "family": "qwen3.6",
1277
1415
  "attachment": true,
1278
1416
  "reasoning": true,
1279
1417
  "reasoning_options": [
@@ -1287,41 +1425,41 @@
1287
1425
  "tool_call": true,
1288
1426
  "structured_output": true,
1289
1427
  "temperature": true,
1290
- "release_date": "2026-04-17",
1291
- "last_updated": "2026-04-17",
1428
+ "release_date": "2026-04-27",
1429
+ "last_updated": "2026-04-27",
1292
1430
  "modalities": {
1293
1431
  "input": [
1294
1432
  "text",
1295
1433
  "image",
1296
- "video",
1297
- "audio"
1434
+ "video"
1298
1435
  ],
1299
1436
  "output": [
1300
1437
  "text"
1301
1438
  ]
1302
1439
  },
1303
- "open_weights": true,
1440
+ "open_weights": false,
1304
1441
  "limit": {
1305
- "context": 262144,
1442
+ "context": 1000000,
1306
1443
  "output": 65536
1307
1444
  },
1308
1445
  "cost": {
1309
- "input": 0.248,
1310
- "output": 1.485
1446
+ "input": 0.1875,
1447
+ "output": 1.125,
1448
+ "cache_write": 0.234375
1311
1449
  }
1312
1450
  },
1313
- "qwen3-max": {
1314
- "id": "qwen3-max",
1315
- "name": "Qwen3 Max",
1316
- "description": "Flagship Qwen3 model for coding agents, complex reasoning, and tool use",
1451
+ "qwen3-coder-flash": {
1452
+ "id": "qwen3-coder-flash",
1453
+ "name": "Qwen3 Coder Flash",
1454
+ "description": "Qwen coding model for software agents, repository edits, and code reasoning",
1317
1455
  "family": "qwen",
1318
1456
  "attachment": false,
1319
1457
  "reasoning": false,
1320
1458
  "tool_call": true,
1321
1459
  "temperature": true,
1322
1460
  "knowledge": "2025-04",
1323
- "release_date": "2025-09-23",
1324
- "last_updated": "2025-09-23",
1461
+ "release_date": "2025-07-28",
1462
+ "last_updated": "2025-07-28",
1325
1463
  "modalities": {
1326
1464
  "input": [
1327
1465
  "text"
@@ -1332,34 +1470,61 @@
1332
1470
  },
1333
1471
  "open_weights": false,
1334
1472
  "limit": {
1335
- "context": 262144,
1473
+ "context": 1000000,
1336
1474
  "output": 65536
1337
1475
  },
1338
1476
  "cost": {
1339
- "input": 1.2,
1340
- "output": 6
1477
+ "input": 0.3,
1478
+ "output": 1.5
1341
1479
  }
1342
1480
  },
1343
- "qwen-plus": {
1344
- "id": "qwen-plus",
1345
- "name": "Qwen Plus",
1346
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1481
+ "qwen2-5-omni-7b": {
1482
+ "id": "qwen2-5-omni-7b",
1483
+ "name": "Qwen2.5-Omni 7B",
1484
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1347
1485
  "family": "qwen",
1348
1486
  "attachment": false,
1349
- "reasoning": true,
1350
- "reasoning_options": [
1351
- {
1352
- "type": "toggle"
1353
- },
1354
- {
1355
- "type": "budget_tokens"
1356
- }
1357
- ],
1487
+ "reasoning": false,
1358
1488
  "tool_call": true,
1359
1489
  "temperature": true,
1360
1490
  "knowledge": "2024-04",
1361
- "release_date": "2024-01-25",
1362
- "last_updated": "2025-09-11",
1491
+ "release_date": "2024-12",
1492
+ "last_updated": "2024-12",
1493
+ "modalities": {
1494
+ "input": [
1495
+ "text",
1496
+ "image",
1497
+ "audio",
1498
+ "video"
1499
+ ],
1500
+ "output": [
1501
+ "text",
1502
+ "audio"
1503
+ ]
1504
+ },
1505
+ "open_weights": true,
1506
+ "limit": {
1507
+ "context": 32768,
1508
+ "output": 2048
1509
+ },
1510
+ "cost": {
1511
+ "input": 0.1,
1512
+ "output": 0.4,
1513
+ "input_audio": 6.76
1514
+ }
1515
+ },
1516
+ "qwen-mt-plus": {
1517
+ "id": "qwen-mt-plus",
1518
+ "name": "Qwen-MT Plus",
1519
+ "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
1520
+ "family": "qwen",
1521
+ "attachment": false,
1522
+ "reasoning": false,
1523
+ "tool_call": false,
1524
+ "temperature": true,
1525
+ "knowledge": "2024-04",
1526
+ "release_date": "2025-01",
1527
+ "last_updated": "2025-01",
1363
1528
  "modalities": {
1364
1529
  "input": [
1365
1530
  "text"
@@ -1370,41 +1535,35 @@
1370
1535
  },
1371
1536
  "open_weights": false,
1372
1537
  "limit": {
1373
- "context": 1000000,
1374
- "output": 32768
1538
+ "context": 16384,
1539
+ "output": 8192
1375
1540
  },
1376
1541
  "cost": {
1377
- "input": 0.4,
1378
- "output": 1.2,
1379
- "reasoning": 4
1542
+ "input": 2.46,
1543
+ "output": 7.37
1380
1544
  }
1381
1545
  },
1382
- "qwen3.5-122b-a10b": {
1383
- "id": "qwen3.5-122b-a10b",
1384
- "name": "Qwen3.5 122B-A10B",
1546
+ "qwen3-vl-235b-a22b": {
1547
+ "id": "qwen3-vl-235b-a22b",
1548
+ "name": "Qwen3-VL 235B-A22B",
1385
1549
  "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1386
1550
  "family": "qwen",
1387
- "attachment": true,
1551
+ "attachment": false,
1388
1552
  "reasoning": true,
1389
1553
  "reasoning_options": [
1390
- {
1391
- "type": "toggle"
1392
- },
1393
1554
  {
1394
1555
  "type": "budget_tokens"
1395
1556
  }
1396
1557
  ],
1397
1558
  "tool_call": true,
1398
- "structured_output": true,
1399
1559
  "temperature": true,
1400
- "release_date": "2026-02-23",
1401
- "last_updated": "2026-02-23",
1560
+ "knowledge": "2025-04",
1561
+ "release_date": "2025-04",
1562
+ "last_updated": "2025-04",
1402
1563
  "modalities": {
1403
1564
  "input": [
1404
1565
  "text",
1405
- "image",
1406
- "video",
1407
- "audio"
1566
+ "image"
1408
1567
  ],
1409
1568
  "output": [
1410
1569
  "text"
@@ -1412,54 +1571,49 @@
1412
1571
  },
1413
1572
  "open_weights": true,
1414
1573
  "limit": {
1415
- "context": 262144,
1416
- "output": 65536
1574
+ "context": 131072,
1575
+ "output": 32768
1417
1576
  },
1418
1577
  "cost": {
1419
- "input": 0.4,
1420
- "output": 3.2
1578
+ "input": 0.7,
1579
+ "output": 2.8,
1580
+ "reasoning": 8.4
1421
1581
  }
1422
1582
  },
1423
- "qwen3-omni-flash-realtime": {
1424
- "id": "qwen3-omni-flash-realtime",
1425
- "name": "Qwen3-Omni Flash Realtime",
1426
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1583
+ "qwen3-coder-plus": {
1584
+ "id": "qwen3-coder-plus",
1585
+ "name": "Qwen3 Coder Plus",
1586
+ "description": "Hosted Qwen coder for software agents, repo edits, and long-context code",
1427
1587
  "family": "qwen",
1428
1588
  "attachment": false,
1429
1589
  "reasoning": false,
1430
1590
  "tool_call": true,
1431
1591
  "temperature": true,
1432
- "knowledge": "2024-04",
1433
- "release_date": "2025-09-15",
1434
- "last_updated": "2025-09-15",
1592
+ "knowledge": "2025-04",
1593
+ "release_date": "2025-07-23",
1594
+ "last_updated": "2025-07-23",
1435
1595
  "modalities": {
1436
1596
  "input": [
1437
- "text",
1438
- "image",
1439
- "audio",
1440
- "video"
1597
+ "text"
1441
1598
  ],
1442
1599
  "output": [
1443
- "text",
1444
- "audio"
1600
+ "text"
1445
1601
  ]
1446
1602
  },
1447
- "open_weights": false,
1603
+ "open_weights": true,
1448
1604
  "limit": {
1449
- "context": 65536,
1450
- "output": 16384
1605
+ "context": 1048576,
1606
+ "output": 65536
1451
1607
  },
1452
1608
  "cost": {
1453
- "input": 0.52,
1454
- "output": 1.99,
1455
- "input_audio": 4.57,
1456
- "output_audio": 18.13
1609
+ "input": 1,
1610
+ "output": 5
1457
1611
  }
1458
1612
  },
1459
- "qwen2-5-vl-7b-instruct": {
1460
- "id": "qwen2-5-vl-7b-instruct",
1461
- "name": "Qwen2.5-VL 7B Instruct",
1462
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1613
+ "qwen2-5-14b-instruct": {
1614
+ "id": "qwen2-5-14b-instruct",
1615
+ "name": "Qwen2.5 14B Instruct",
1616
+ "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1463
1617
  "family": "qwen",
1464
1618
  "attachment": false,
1465
1619
  "reasoning": false,
@@ -1470,8 +1624,7 @@
1470
1624
  "last_updated": "2024-09",
1471
1625
  "modalities": {
1472
1626
  "input": [
1473
- "text",
1474
- "image"
1627
+ "text"
1475
1628
  ],
1476
1629
  "output": [
1477
1630
  "text"
@@ -1484,48 +1637,7 @@
1484
1637
  },
1485
1638
  "cost": {
1486
1639
  "input": 0.35,
1487
- "output": 1.05
1488
- }
1489
- },
1490
- "qwen3.6-flash": {
1491
- "id": "qwen3.6-flash",
1492
- "name": "Qwen3.6 Flash",
1493
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1494
- "family": "qwen3.6",
1495
- "attachment": true,
1496
- "reasoning": true,
1497
- "reasoning_options": [
1498
- {
1499
- "type": "toggle"
1500
- },
1501
- {
1502
- "type": "budget_tokens"
1503
- }
1504
- ],
1505
- "tool_call": true,
1506
- "structured_output": true,
1507
- "temperature": true,
1508
- "release_date": "2026-04-27",
1509
- "last_updated": "2026-04-27",
1510
- "modalities": {
1511
- "input": [
1512
- "text",
1513
- "image",
1514
- "video"
1515
- ],
1516
- "output": [
1517
- "text"
1518
- ]
1519
- },
1520
- "open_weights": false,
1521
- "limit": {
1522
- "context": 1000000,
1523
- "output": 65536
1524
- },
1525
- "cost": {
1526
- "input": 0.1875,
1527
- "output": 1.125,
1528
- "cache_write": 0.234375
1640
+ "output": 1.4
1529
1641
  }
1530
1642
  },
1531
1643
  "qwen3.8-flash": {
@@ -1579,23 +1691,66 @@
1579
1691
  "output": 0.47,
1580
1692
  "cache_read": 0.016,
1581
1693
  "cache_write": 0.2
1694
+ },
1695
+ "canonical_model_id": "alibaba/qwen3.8-flash"
1696
+ },
1697
+ "qwen2-5-vl-7b-instruct": {
1698
+ "id": "qwen2-5-vl-7b-instruct",
1699
+ "name": "Qwen2.5-VL 7B Instruct",
1700
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1701
+ "family": "qwen",
1702
+ "attachment": false,
1703
+ "reasoning": false,
1704
+ "tool_call": true,
1705
+ "temperature": true,
1706
+ "knowledge": "2024-04",
1707
+ "release_date": "2024-09",
1708
+ "last_updated": "2024-09",
1709
+ "modalities": {
1710
+ "input": [
1711
+ "text",
1712
+ "image"
1713
+ ],
1714
+ "output": [
1715
+ "text"
1716
+ ]
1717
+ },
1718
+ "open_weights": true,
1719
+ "limit": {
1720
+ "context": 131072,
1721
+ "output": 8192
1722
+ },
1723
+ "cost": {
1724
+ "input": 0.35,
1725
+ "output": 1.05
1582
1726
  }
1583
1727
  },
1584
- "qwen2-5-14b-instruct": {
1585
- "id": "qwen2-5-14b-instruct",
1586
- "name": "Qwen2.5 14B Instruct",
1587
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1728
+ "qwen3.5-35b-a3b": {
1729
+ "id": "qwen3.5-35b-a3b",
1730
+ "name": "Qwen3.5 35B-A3B",
1731
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1588
1732
  "family": "qwen",
1589
- "attachment": false,
1590
- "reasoning": false,
1733
+ "attachment": true,
1734
+ "reasoning": true,
1735
+ "reasoning_options": [
1736
+ {
1737
+ "type": "toggle"
1738
+ },
1739
+ {
1740
+ "type": "budget_tokens"
1741
+ }
1742
+ ],
1591
1743
  "tool_call": true,
1744
+ "structured_output": true,
1592
1745
  "temperature": true,
1593
- "knowledge": "2024-04",
1594
- "release_date": "2024-09",
1595
- "last_updated": "2024-09",
1746
+ "release_date": "2026-02-23",
1747
+ "last_updated": "2026-02-23",
1596
1748
  "modalities": {
1597
1749
  "input": [
1598
- "text"
1750
+ "text",
1751
+ "image",
1752
+ "video",
1753
+ "audio"
1599
1754
  ],
1600
1755
  "output": [
1601
1756
  "text"
@@ -1603,12 +1758,12 @@
1603
1758
  },
1604
1759
  "open_weights": true,
1605
1760
  "limit": {
1606
- "context": 131072,
1607
- "output": 8192
1761
+ "context": 262144,
1762
+ "output": 65536
1608
1763
  },
1609
1764
  "cost": {
1610
- "input": 0.35,
1611
- "output": 1.4
1765
+ "input": 0.25,
1766
+ "output": 2
1612
1767
  }
1613
1768
  },
1614
1769
  "qwen3.6-max-preview": {
@@ -1651,18 +1806,18 @@
1651
1806
  "cache_write": 1.625
1652
1807
  }
1653
1808
  },
1654
- "qwen-plus-character-ja": {
1655
- "id": "qwen-plus-character-ja",
1656
- "name": "Qwen Plus Character (Japanese)",
1657
- "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1809
+ "qwen-mt-turbo": {
1810
+ "id": "qwen-mt-turbo",
1811
+ "name": "Qwen-MT Turbo",
1812
+ "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
1658
1813
  "family": "qwen",
1659
1814
  "attachment": false,
1660
1815
  "reasoning": false,
1661
- "tool_call": true,
1816
+ "tool_call": false,
1662
1817
  "temperature": true,
1663
1818
  "knowledge": "2024-04",
1664
- "release_date": "2024-01",
1665
- "last_updated": "2024-01",
1819
+ "release_date": "2025-01",
1820
+ "last_updated": "2025-01",
1666
1821
  "modalities": {
1667
1822
  "input": [
1668
1823
  "text"
@@ -1673,18 +1828,18 @@
1673
1828
  },
1674
1829
  "open_weights": false,
1675
1830
  "limit": {
1676
- "context": 8192,
1677
- "output": 512
1831
+ "context": 16384,
1832
+ "output": 8192
1678
1833
  },
1679
1834
  "cost": {
1680
- "input": 0.5,
1681
- "output": 1.4
1835
+ "input": 0.16,
1836
+ "output": 0.49
1682
1837
  }
1683
1838
  },
1684
- "qwen3.8-max": {
1685
- "id": "qwen3.8-max",
1686
- "name": "Qwen3.8 Max",
1687
- "description": "2.4-trillion-parameter MoE flagship for coding, professional work, multimodal understanding, and long-horizon agentic workflows",
1839
+ "qwen3.5-397b-a17b": {
1840
+ "id": "qwen3.5-397b-a17b",
1841
+ "name": "Qwen3.5 397B-A17B",
1842
+ "description": "Large open Qwen multimodal MoE for visual agents and long technical tasks",
1688
1843
  "family": "qwen",
1689
1844
  "attachment": true,
1690
1845
  "reasoning": true,
@@ -1693,48 +1848,33 @@
1693
1848
  "type": "toggle"
1694
1849
  },
1695
1850
  {
1696
- "type": "effort",
1697
- "values": [
1698
- "low",
1699
- "medium",
1700
- "xhigh"
1701
- ]
1702
- },
1703
- {
1704
- "type": "budget_tokens",
1705
- "min": 0,
1706
- "max": 262144
1851
+ "type": "budget_tokens"
1707
1852
  }
1708
1853
  ],
1709
1854
  "tool_call": true,
1710
- "interleaved": {
1711
- "field": "reasoning_content"
1712
- },
1713
1855
  "structured_output": true,
1714
1856
  "temperature": true,
1715
- "release_date": "2026-08-03",
1716
- "last_updated": "2026-08-03",
1857
+ "release_date": "2026-02-15",
1858
+ "last_updated": "2026-02-15",
1717
1859
  "modalities": {
1718
1860
  "input": [
1719
1861
  "text",
1720
1862
  "image",
1721
1863
  "video",
1722
- "pdf"
1864
+ "audio"
1723
1865
  ],
1724
1866
  "output": [
1725
1867
  "text"
1726
1868
  ]
1727
1869
  },
1728
- "open_weights": false,
1870
+ "open_weights": true,
1729
1871
  "limit": {
1730
- "context": 1000000,
1731
- "output": 131072
1872
+ "context": 262144,
1873
+ "output": 65536
1732
1874
  },
1733
1875
  "cost": {
1734
- "input": 2,
1735
- "output": 6,
1736
- "cache_read": 0.25,
1737
- "cache_write": 2.5
1876
+ "input": 0.6,
1877
+ "output": 3.6
1738
1878
  }
1739
1879
  },
1740
1880
  "qwen3-8b": {
@@ -1776,12 +1916,164 @@
1776
1916
  "reasoning": 2.1
1777
1917
  }
1778
1918
  },
1779
- "qwen3-235b-a22b": {
1780
- "id": "qwen3-235b-a22b",
1781
- "name": "Qwen3 235B-A22B",
1782
- "description": "Large open Qwen MoE for multilingual reasoning, coding, and tool use",
1919
+ "glm-5.2": {
1920
+ "id": "glm-5.2",
1921
+ "name": "GLM-5.2",
1922
+ "description": "Open flagship GLM for long-horizon coding agents and million-token context work",
1923
+ "family": "glm",
1924
+ "attachment": false,
1925
+ "reasoning": true,
1926
+ "reasoning_options": [
1927
+ {
1928
+ "type": "effort",
1929
+ "values": [
1930
+ "none",
1931
+ "minimal",
1932
+ "low",
1933
+ "medium",
1934
+ "high",
1935
+ "xhigh",
1936
+ "max"
1937
+ ]
1938
+ }
1939
+ ],
1940
+ "tool_call": true,
1941
+ "interleaved": {
1942
+ "field": "reasoning_content"
1943
+ },
1944
+ "structured_output": true,
1945
+ "temperature": true,
1946
+ "release_date": "2026-06-13",
1947
+ "last_updated": "2026-06-13",
1948
+ "modalities": {
1949
+ "input": [
1950
+ "text"
1951
+ ],
1952
+ "output": [
1953
+ "text"
1954
+ ]
1955
+ },
1956
+ "open_weights": true,
1957
+ "limit": {
1958
+ "context": 1000000,
1959
+ "output": 131072
1960
+ },
1961
+ "cost": {
1962
+ "input": 1.4,
1963
+ "output": 4.4,
1964
+ "cache_read": 0.28,
1965
+ "cache_write": 0
1966
+ },
1967
+ "canonical_model_id": "zhipuai/glm-5.2"
1968
+ },
1969
+ "qwen-omni-turbo": {
1970
+ "id": "qwen-omni-turbo",
1971
+ "name": "Qwen-Omni Turbo",
1972
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1973
+ "family": "qwen",
1974
+ "attachment": false,
1975
+ "reasoning": false,
1976
+ "tool_call": true,
1977
+ "temperature": true,
1978
+ "knowledge": "2024-04",
1979
+ "release_date": "2025-01-19",
1980
+ "last_updated": "2025-03-26",
1981
+ "modalities": {
1982
+ "input": [
1983
+ "text",
1984
+ "image",
1985
+ "audio",
1986
+ "video"
1987
+ ],
1988
+ "output": [
1989
+ "text",
1990
+ "audio"
1991
+ ]
1992
+ },
1993
+ "open_weights": false,
1994
+ "limit": {
1995
+ "context": 32768,
1996
+ "output": 2048
1997
+ },
1998
+ "cost": {
1999
+ "input": 0.07,
2000
+ "output": 0.27,
2001
+ "input_audio": 4.44,
2002
+ "output_audio": 8.89
2003
+ }
2004
+ },
2005
+ "qwen3-omni-flash-realtime": {
2006
+ "id": "qwen3-omni-flash-realtime",
2007
+ "name": "Qwen3-Omni Flash Realtime",
2008
+ "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
2009
+ "family": "qwen",
2010
+ "attachment": false,
2011
+ "reasoning": false,
2012
+ "tool_call": true,
2013
+ "temperature": true,
2014
+ "knowledge": "2024-04",
2015
+ "release_date": "2025-09-15",
2016
+ "last_updated": "2025-09-15",
2017
+ "modalities": {
2018
+ "input": [
2019
+ "text",
2020
+ "image",
2021
+ "audio",
2022
+ "video"
2023
+ ],
2024
+ "output": [
2025
+ "text",
2026
+ "audio"
2027
+ ]
2028
+ },
2029
+ "open_weights": false,
2030
+ "limit": {
2031
+ "context": 65536,
2032
+ "output": 16384
2033
+ },
2034
+ "cost": {
2035
+ "input": 0.52,
2036
+ "output": 1.99,
2037
+ "input_audio": 4.57,
2038
+ "output_audio": 18.13
2039
+ }
2040
+ },
2041
+ "qwen3-asr-flash": {
2042
+ "id": "qwen3-asr-flash",
2043
+ "name": "Qwen3-ASR Flash",
2044
+ "description": "Speech transcription model for accurate audio-to-text and captioning workflows",
2045
+ "family": "qwen",
2046
+ "attachment": false,
2047
+ "reasoning": false,
2048
+ "tool_call": false,
2049
+ "temperature": false,
2050
+ "knowledge": "2024-04",
2051
+ "release_date": "2025-09-08",
2052
+ "last_updated": "2025-09-08",
2053
+ "modalities": {
2054
+ "input": [
2055
+ "audio"
2056
+ ],
2057
+ "output": [
2058
+ "text"
2059
+ ]
2060
+ },
2061
+ "open_weights": false,
2062
+ "limit": {
2063
+ "context": 53248,
2064
+ "output": 4096
2065
+ },
2066
+ "cost": {
2067
+ "input": 0.035,
2068
+ "output": 0.035
2069
+ }
2070
+ },
2071
+ "qwen3.6-27b": {
2072
+ "id": "qwen3.6-27b",
2073
+ "name": "Qwen3.6 27B",
2074
+ "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1783
2075
  "family": "qwen",
1784
- "attachment": false,
2076
+ "attachment": true,
1785
2077
  "reasoning": true,
1786
2078
  "reasoning_options": [
1787
2079
  {
@@ -1792,13 +2084,16 @@
1792
2084
  }
1793
2085
  ],
1794
2086
  "tool_call": true,
2087
+ "structured_output": true,
1795
2088
  "temperature": true,
1796
- "knowledge": "2025-04",
1797
- "release_date": "2025-04",
1798
- "last_updated": "2025-04",
2089
+ "release_date": "2026-04-22",
2090
+ "last_updated": "2026-04-22",
1799
2091
  "modalities": {
1800
2092
  "input": [
1801
- "text"
2093
+ "text",
2094
+ "image",
2095
+ "video",
2096
+ "audio"
1802
2097
  ],
1803
2098
  "output": [
1804
2099
  "text"
@@ -1806,19 +2101,18 @@
1806
2101
  },
1807
2102
  "open_weights": true,
1808
2103
  "limit": {
1809
- "context": 131072,
1810
- "output": 16384
2104
+ "context": 262144,
2105
+ "output": 65536
1811
2106
  },
1812
2107
  "cost": {
1813
- "input": 0.7,
1814
- "output": 2.8,
1815
- "reasoning": 8.4
2108
+ "input": 0.6,
2109
+ "output": 3.6
1816
2110
  }
1817
2111
  },
1818
- "qwen3.7-plus": {
1819
- "id": "qwen3.7-plus",
1820
- "name": "Qwen3.7 Plus",
1821
- "description": "Multimodal Qwen workhorse for long-context agents, visual inputs, and coding",
2112
+ "qwen3.6-plus": {
2113
+ "id": "qwen3.6-plus",
2114
+ "name": "Qwen3.6 Plus",
2115
+ "description": "Earlier Qwen multimodal workhorse for million-token agent and document tasks",
1822
2116
  "family": "qwen",
1823
2117
  "attachment": false,
1824
2118
  "reasoning": true,
@@ -1833,8 +2127,8 @@
1833
2127
  "tool_call": true,
1834
2128
  "temperature": true,
1835
2129
  "knowledge": "2025-04",
1836
- "release_date": "2026-06-02",
1837
- "last_updated": "2026-06-04",
2130
+ "release_date": "2026-04-02",
2131
+ "last_updated": "2026-04-02",
1838
2132
  "modalities": {
1839
2133
  "input": [
1840
2134
  "text",
@@ -1848,7 +2142,7 @@
1848
2142
  "open_weights": false,
1849
2143
  "limit": {
1850
2144
  "context": 1000000,
1851
- "output": 131072
2145
+ "output": 65536
1852
2146
  },
1853
2147
  "cost": {
1854
2148
  "input": 0.5,
@@ -1875,128 +2169,17 @@
1875
2169
  }
1876
2170
  }
1877
2171
  },
1878
- "qwen-omni-turbo": {
1879
- "id": "qwen-omni-turbo",
1880
- "name": "Qwen-Omni Turbo",
1881
- "description": "Qwen omni model for text, vision, audio, and multimodal agent tasks",
1882
- "family": "qwen",
1883
- "attachment": false,
1884
- "reasoning": false,
1885
- "tool_call": true,
1886
- "temperature": true,
1887
- "knowledge": "2024-04",
1888
- "release_date": "2025-01-19",
1889
- "last_updated": "2025-03-26",
1890
- "modalities": {
1891
- "input": [
1892
- "text",
1893
- "image",
1894
- "audio",
1895
- "video"
1896
- ],
1897
- "output": [
1898
- "text",
1899
- "audio"
1900
- ]
1901
- },
1902
- "open_weights": false,
1903
- "limit": {
1904
- "context": 32768,
1905
- "output": 2048
1906
- },
1907
- "cost": {
1908
- "input": 0.07,
1909
- "output": 0.27,
1910
- "input_audio": 4.44,
1911
- "output_audio": 8.89
1912
- }
1913
- },
1914
- "qwen3-coder-480b-a35b-instruct": {
1915
- "id": "qwen3-coder-480b-a35b-instruct",
1916
- "name": "Qwen3-Coder 480B-A35B Instruct",
1917
- "description": "Open Qwen coding heavyweight for repository reasoning and agentic engineering",
1918
- "family": "qwen",
1919
- "attachment": false,
1920
- "reasoning": false,
1921
- "tool_call": true,
1922
- "temperature": true,
1923
- "knowledge": "2025-04",
1924
- "release_date": "2025-04",
1925
- "last_updated": "2025-04",
1926
- "modalities": {
1927
- "input": [
1928
- "text"
1929
- ],
1930
- "output": [
1931
- "text"
1932
- ]
1933
- },
1934
- "open_weights": true,
1935
- "limit": {
1936
- "context": 262144,
1937
- "output": 65536
1938
- },
1939
- "cost": {
1940
- "input": 1.5,
1941
- "output": 7.5,
1942
- "tiers": [
1943
- {
1944
- "input": 2.7,
1945
- "output": 13.5,
1946
- "tier": {
1947
- "type": "context",
1948
- "size": 32000
1949
- }
1950
- },
1951
- {
1952
- "input": 4.5,
1953
- "output": 22.5,
1954
- "tier": {
1955
- "type": "context",
1956
- "size": 128000
1957
- }
1958
- }
1959
- ]
1960
- }
1961
- },
1962
- "qwen2-5-32b-instruct": {
1963
- "id": "qwen2-5-32b-instruct",
1964
- "name": "Qwen2.5 32B Instruct",
2172
+ "qwen3-14b": {
2173
+ "id": "qwen3-14b",
2174
+ "name": "Qwen3 14B",
1965
2175
  "description": "Qwen instruction model for multilingual chat, reasoning, and tool use",
1966
2176
  "family": "qwen",
1967
2177
  "attachment": false,
1968
- "reasoning": false,
1969
- "tool_call": true,
1970
- "temperature": true,
1971
- "knowledge": "2024-04",
1972
- "release_date": "2024-09",
1973
- "last_updated": "2024-09",
1974
- "modalities": {
1975
- "input": [
1976
- "text"
1977
- ],
1978
- "output": [
1979
- "text"
1980
- ]
1981
- },
1982
- "open_weights": true,
1983
- "limit": {
1984
- "context": 131072,
1985
- "output": 8192
1986
- },
1987
- "cost": {
1988
- "input": 0.7,
1989
- "output": 2.8
1990
- }
1991
- },
1992
- "qwen3-vl-235b-a22b": {
1993
- "id": "qwen3-vl-235b-a22b",
1994
- "name": "Qwen3-VL 235B-A22B",
1995
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
1996
- "family": "qwen",
1997
- "attachment": false,
1998
2178
  "reasoning": true,
1999
2179
  "reasoning_options": [
2180
+ {
2181
+ "type": "toggle"
2182
+ },
2000
2183
  {
2001
2184
  "type": "budget_tokens"
2002
2185
  }
@@ -2008,8 +2191,7 @@
2008
2191
  "last_updated": "2025-04",
2009
2192
  "modalities": {
2010
2193
  "input": [
2011
- "text",
2012
- "image"
2194
+ "text"
2013
2195
  ],
2014
2196
  "output": [
2015
2197
  "text"
@@ -2018,20 +2200,20 @@
2018
2200
  "open_weights": true,
2019
2201
  "limit": {
2020
2202
  "context": 131072,
2021
- "output": 32768
2203
+ "output": 8192
2022
2204
  },
2023
2205
  "cost": {
2024
- "input": 0.7,
2025
- "output": 2.8,
2026
- "reasoning": 8.4
2206
+ "input": 0.35,
2207
+ "output": 1.4,
2208
+ "reasoning": 4.2
2027
2209
  }
2028
2210
  },
2029
- "qwen3.5-plus": {
2030
- "id": "qwen3.5-plus",
2031
- "name": "Qwen3.5 Plus",
2032
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
2211
+ "qwen3.7-plus": {
2212
+ "id": "qwen3.7-plus",
2213
+ "name": "Qwen3.7 Plus",
2214
+ "description": "Multimodal Qwen workhorse for long-context agents, visual inputs, and coding",
2033
2215
  "family": "qwen",
2034
- "attachment": false,
2216
+ "attachment": true,
2035
2217
  "reasoning": true,
2036
2218
  "reasoning_options": [
2037
2219
  {
@@ -2042,10 +2224,14 @@
2042
2224
  }
2043
2225
  ],
2044
2226
  "tool_call": true,
2227
+ "interleaved": {
2228
+ "field": "reasoning_content"
2229
+ },
2230
+ "structured_output": true,
2045
2231
  "temperature": true,
2046
2232
  "knowledge": "2025-04",
2047
- "release_date": "2026-02-16",
2048
- "last_updated": "2026-02-16",
2233
+ "release_date": "2026-06-02",
2234
+ "last_updated": "2026-06-04",
2049
2235
  "modalities": {
2050
2236
  "input": [
2051
2237
  "text",
@@ -2059,57 +2245,45 @@
2059
2245
  "open_weights": false,
2060
2246
  "limit": {
2061
2247
  "context": 1000000,
2062
- "output": 65536
2248
+ "output": 131072
2063
2249
  },
2064
2250
  "cost": {
2065
2251
  "input": 0.4,
2066
- "output": 2.4,
2067
- "reasoning": 2.4
2252
+ "output": 1.6,
2253
+ "cache_read": 0.04,
2254
+ "cache_write": 0.5,
2255
+ "tiers": [
2256
+ {
2257
+ "input": 1.2,
2258
+ "output": 4.8,
2259
+ "cache_read": 0.12,
2260
+ "cache_write": 1.5,
2261
+ "tier": {
2262
+ "type": "context",
2263
+ "size": 256000
2264
+ }
2265
+ }
2266
+ ],
2267
+ "context_over_200k": {
2268
+ "input": 1.2,
2269
+ "output": 4.8,
2270
+ "cache_read": 0.12,
2271
+ "cache_write": 1.5
2272
+ }
2068
2273
  }
2069
2274
  },
2070
- "qwen-mt-plus": {
2071
- "id": "qwen-mt-plus",
2072
- "name": "Qwen-MT Plus",
2073
- "description": "Translation model for multilingual conversion, localization, and cross-language workflows",
2275
+ "qwen-vl-ocr": {
2276
+ "id": "qwen-vl-ocr",
2277
+ "name": "Qwen-VL OCR",
2278
+ "description": "OCR model for extracting structured text from documents and screenshots",
2074
2279
  "family": "qwen",
2075
2280
  "attachment": false,
2076
2281
  "reasoning": false,
2077
2282
  "tool_call": false,
2078
2283
  "temperature": true,
2079
2284
  "knowledge": "2024-04",
2080
- "release_date": "2025-01",
2081
- "last_updated": "2025-01",
2082
- "modalities": {
2083
- "input": [
2084
- "text"
2085
- ],
2086
- "output": [
2087
- "text"
2088
- ]
2089
- },
2090
- "open_weights": false,
2091
- "limit": {
2092
- "context": 16384,
2093
- "output": 8192
2094
- },
2095
- "cost": {
2096
- "input": 2.46,
2097
- "output": 7.37
2098
- }
2099
- },
2100
- "qvq-max": {
2101
- "id": "qvq-max",
2102
- "name": "QVQ Max",
2103
- "description": "Qwen vision-language model for visual reasoning, documents, and agent tasks",
2104
- "family": "qvq",
2105
- "attachment": false,
2106
- "reasoning": true,
2107
- "reasoning_options": [],
2108
- "tool_call": true,
2109
- "temperature": true,
2110
- "knowledge": "2024-04",
2111
- "release_date": "2025-03-25",
2112
- "last_updated": "2025-03-25",
2285
+ "release_date": "2024-10-28",
2286
+ "last_updated": "2025-04-13",
2113
2287
  "modalities": {
2114
2288
  "input": [
2115
2289
  "text",
@@ -2121,12 +2295,12 @@
2121
2295
  },
2122
2296
  "open_weights": false,
2123
2297
  "limit": {
2124
- "context": 131072,
2125
- "output": 8192
2298
+ "context": 34096,
2299
+ "output": 4096
2126
2300
  },
2127
2301
  "cost": {
2128
- "input": 1.2,
2129
- "output": 4.8
2302
+ "input": 0.72,
2303
+ "output": 0.72
2130
2304
  }
2131
2305
  }
2132
2306
  }