@lobehub/lobehub 2.0.0-next.101 → 2.0.0-next.103

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 (41) hide show
  1. package/CHANGELOG.md +50 -0
  2. package/changelog/v1.json +18 -0
  3. package/package.json +1 -1
  4. package/packages/database/src/models/aiModel.ts +1 -0
  5. package/packages/model-bank/package.json +1 -0
  6. package/packages/model-bank/src/aiModels/aihubmix.ts +29 -0
  7. package/packages/model-bank/src/aiModels/google.ts +72 -10
  8. package/packages/model-bank/src/aiModels/index.ts +3 -0
  9. package/packages/model-bank/src/aiModels/infiniai.ts +5 -22
  10. package/packages/model-bank/src/aiModels/ollamacloud.ts +13 -0
  11. package/packages/model-bank/src/aiModels/siliconcloud.ts +0 -61
  12. package/packages/model-bank/src/aiModels/vertexai.ts +90 -1
  13. package/packages/model-bank/src/aiModels/zenmux.ts +1426 -0
  14. package/packages/model-bank/src/const/modelProvider.ts +1 -0
  15. package/packages/model-bank/src/standard-parameters/index.ts +9 -0
  16. package/packages/model-bank/src/types/aiModel.ts +1 -0
  17. package/packages/model-runtime/src/core/openaiCompatibleFactory/index.test.ts +2 -2
  18. package/packages/model-runtime/src/core/streams/google/index.ts +7 -2
  19. package/packages/model-runtime/src/core/streams/openai/__snapshots__/responsesStream.test.ts.snap +166 -166
  20. package/packages/model-runtime/src/index.ts +1 -1
  21. package/packages/model-runtime/src/providers/google/createImage.ts +1 -0
  22. package/packages/model-runtime/src/providers/google/index.ts +11 -1
  23. package/packages/model-runtime/src/providers/zenmux/index.test.ts +320 -0
  24. package/packages/model-runtime/src/providers/zenmux/index.ts +84 -0
  25. package/packages/model-runtime/src/runtimeMap.ts +2 -0
  26. package/packages/types/src/user/settings/keyVaults.ts +1 -0
  27. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/components/ModelSelect/ImageModelItem.tsx +16 -1
  28. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/components/ModelSelect/index.tsx +10 -9
  29. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/components/ResolutionSelect.tsx +88 -0
  30. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/index.tsx +9 -0
  31. package/src/app/[variants]/(main)/settings/hooks/useCategory.tsx +3 -3
  32. package/src/app/[variants]/(main)/settings/provider/features/ModelList/ModelItem.tsx +2 -7
  33. package/src/components/ModelSelect/NewModelBadge.tsx +23 -0
  34. package/src/components/ModelSelect/index.tsx +4 -0
  35. package/src/config/modelProviders/index.ts +3 -0
  36. package/src/config/modelProviders/zenmux.ts +21 -0
  37. package/src/envs/llm.ts +6 -0
  38. package/src/locales/default/image.ts +8 -0
  39. package/src/store/aiInfra/slices/aiProvider/action.ts +3 -0
  40. package/src/store/chat/slices/aiChat/actions/__tests__/conversationLifecycle.test.ts +3 -0
  41. package/src/store/chat/slices/aiChat/actions/streamingExecutor.ts +11 -0
package/packages/model-bank/src/aiModels/zenmux.ts ADDED
@@ -0,0 +1,1426 @@
1
+ import { AIChatModelCard } from '../types/aiModel';
2
+
3
+ const zenmuxChatModels: AIChatModelCard[] = [
4
+ {
5
+ description:
6
+ 'ZenMux 的自动路由功能会根据你的请求内容，在支持的模型中自动选择当前性价比最高、表现最好的模型。',
7
+ displayName: 'Auto Router',
8
+ id: 'zenmux/auto',
9
+ type: 'chat',
10
+ },
11
+ {
12
+ abilities: {
13
+ functionCall: true,
14
+ imageOutput: true,
15
+ reasoning: true,
16
+ vision: true,
17
+ },
18
+ contextWindowTokens: 65_536,
19
+ description:
20
+ 'Gemini 3 Pro Image（Nano Banana Pro）是 Google 的图像生成模型，同时支持多模态对话。',
21
+ displayName: 'Gemini 3 Pro Image (Nano Banana Pro)',
22
+ enabled: true,
23
+ id: 'google/gemini-3-pro-image-preview',
24
+ maxOutput: 32_768,
25
+ pricing: {
26
+ units: [
27
+ { name: 'imageOutput', rate: 60, strategy: 'fixed', unit: 'millionTokens' },
28
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
29
+ { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
30
+ ],
31
+ },
32
+ releasedAt: '2025-11-20',
33
+ type: 'chat',
34
+ },
35
+ {
36
+ abilities: {
37
+ functionCall: true,
38
+ imageOutput: true,
39
+ reasoning: true,
40
+ vision: true,
41
+ },
42
+ contextWindowTokens: 65_536,
43
+ description: 'Gemini 3 Pro Image 免费版，支持受限额度的多模态生成。',
44
+ displayName: 'Gemini 3 Pro Image (Nano Banana) Free',
45
+ id: 'google/gemini-3-pro-image-preview-free',
46
+ maxOutput: 32_768,
47
+ pricing: {
48
+ units: [
49
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
50
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
51
+ ],
52
+ },
53
+ type: 'chat',
54
+ },
55
+ {
56
+ abilities: {
57
+ functionCall: true,
58
+ reasoning: true,
59
+ vision: true,
60
+ },
61
+ contextWindowTokens: 1_050_000,
62
+ description:
63
+ 'Gemini 3 Pro 是 Gemini 系列下一代多模态推理模型，可理解文本、音频、图像、视频等多种输入，并处理复杂任务与大型代码库。',
64
+ displayName: 'Gemini 3 Pro Preview',
65
+ id: 'google/gemini-3-pro-preview',
66
+ maxOutput: 65_530,
67
+ pricing: {
68
+ units: [
69
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
70
+ { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
71
+ ],
72
+ },
73
+ releasedAt: '2025-11-20',
74
+ type: 'chat',
75
+ },
76
+ {
77
+ abilities: {
78
+ functionCall: true,
79
+ reasoning: true,
80
+ search: true,
81
+ vision: true,
82
+ },
83
+ contextWindowTokens: 1_050_000,
84
+ description:
85
+ 'Gemini 3 Pro 免费预览版，具备与标准版相同的多模态理解与推理能力，但受免费额度与速率限制影响，更适合作为体验与低频使用。',
86
+ displayName: 'Gemini 3 Pro Preview Free',
87
+ id: 'google/gemini-3-pro-preview-free',
88
+ maxOutput: 65_530,
89
+ pricing: {
90
+ units: [
91
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
92
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
93
+ ],
94
+ },
95
+ settings: {
96
+ extendParams: ['thinkingLevel', 'urlContext'],
97
+ searchImpl: 'params',
98
+ searchProvider: 'google',
99
+ },
100
+ type: 'chat',
101
+ },
102
+ {
103
+ abilities: {
104
+ functionCall: true,
105
+ reasoning: true,
106
+ vision: true,
107
+ },
108
+ contextWindowTokens: 400_000,
109
+ description:
110
+ 'GPT-5.1 是 GPT-5 系列最新旗舰模型，相比 GPT-5 在通用推理、指令遵循和对话自然度上均有显著提升，适合广泛任务场景。',
111
+ displayName: 'GPT-5.1',
112
+ id: 'openai/gpt-5.1',
113
+ maxOutput: 128_000,
114
+ pricing: {
115
+ units: [
116
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
117
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
118
+ ],
119
+ },
120
+ settings: {
121
+ extendParams: ['gpt5_1ReasoningEffort', 'textVerbosity'],
122
+ },
123
+ type: 'chat',
124
+ },
125
+ {
126
+ abilities: {
127
+ functionCall: true,
128
+ reasoning: true,
129
+ vision: true,
130
+ },
131
+ contextWindowTokens: 128_000,
132
+ description:
133
+ 'GPT-5.1 Chat 是 GPT-5.1 家族的轻量成员，针对低延迟对话进行优化，同时保留较强的推理与指令执行能力。',
134
+ displayName: 'GPT-5.1 Chat',
135
+ id: 'openai/gpt-5.1-chat',
136
+ maxOutput: 16_380,
137
+ pricing: {
138
+ units: [
139
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
140
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
141
+ ],
142
+ },
143
+ type: 'chat',
144
+ },
145
+ {
146
+ abilities: {
147
+ functionCall: true,
148
+ reasoning: true,
149
+ vision: true,
150
+ },
151
+ contextWindowTokens: 400_000,
152
+ description:
153
+ 'GPT-5.1-Codex 是针对软件工程和编码工作流优化的 GPT-5.1 变体，适合大型重构、复杂调试与长时间自主编码任务。',
154
+ displayName: 'GPT-5.1-Codex',
155
+ id: 'openai/gpt-5.1-codex',
156
+ maxOutput: 128_000,
157
+ pricing: {
158
+ units: [
159
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
160
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
161
+ ],
162
+ },
163
+ settings: {
164
+ extendParams: ['gpt5_1ReasoningEffort'],
165
+ },
166
+ type: 'chat',
167
+ },
168
+ {
169
+ abilities: {
170
+ functionCall: true,
171
+ reasoning: true,
172
+ vision: true,
173
+ },
174
+ contextWindowTokens: 400_000,
175
+ description:
176
+ 'GPT-5.1-Codex-Mini 是 GPT-5.1-Codex 的小型加速版本，更适合对延迟和成本敏感的编码场景。',
177
+ displayName: 'GPT-5.1-Codex-Mini',
178
+ id: 'openai/gpt-5.1-codex-mini',
179
+ maxOutput: 100_000,
180
+ pricing: {
181
+ units: [
182
+ { name: 'textInput', rate: 0.25, strategy: 'fixed', unit: 'millionTokens' },
183
+ { name: 'textOutput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
184
+ ],
185
+ },
186
+ settings: {
187
+ extendParams: ['gpt5_1ReasoningEffort'],
188
+ },
189
+ type: 'chat',
190
+ },
191
+ {
192
+ abilities: {
193
+ functionCall: true,
194
+ reasoning: true,
195
+ vision: true,
196
+ },
197
+ contextWindowTokens: 2_000_000,
198
+ description:
199
+ 'Grok 4 Fast 是 xAI 的高吞吐、低成本模型（支持 2M 上下文窗口），适合需要高并发与长上下文的使用场景。',
200
+ displayName: 'Grok 4.1 Fast',
201
+ id: 'x-ai/grok-4.1-fast',
202
+ maxOutput: 30_000,
203
+ pricing: {
204
+ units: [
205
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
206
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
207
+ ],
208
+ },
209
+ type: 'chat',
210
+ },
211
+ {
212
+ abilities: {
213
+ functionCall: true,
214
+ vision: true,
215
+ },
216
+ contextWindowTokens: 2_000_000,
217
+ description:
218
+ 'Grok 4 Fast（Non-Reasoning）是 xAI 的高吞吐、低成本多模态模型（支持 2M 上下文窗口），面向对延迟和成本敏感但不需要启用模型内推理的场景。它与 Grok 4 Fast 的 reasoning 版本并列，可通过 API 的 reasoning enable 参数在需要时开启推理功能。Prompts 和 completions 可能会被 xAI 或 OpenRouter 用于改进未来模型。',
219
+ displayName: 'Grok 4.1 Fast (Non-Reasoning)',
220
+ id: 'x-ai/grok-4.1-fast-non-reasoning',
221
+ maxOutput: 30_000,
222
+ pricing: {
223
+ units: [
224
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
225
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
226
+ ],
227
+ },
228
+ type: 'chat',
229
+ },
230
+ {
231
+ abilities: {
232
+ functionCall: true,
233
+ reasoning: true,
234
+ vision: true,
235
+ },
236
+ contextWindowTokens: 2_000_000,
237
+ description:
238
+ 'Grok 4 Fast 是 xAI 的高吞吐、低成本模型（支持 2M 上下文窗口），适合需要高并发与长上下文的使用场景。',
239
+ displayName: 'Grok 4 Fast',
240
+ id: 'x-ai/grok-4-fast',
241
+ maxOutput: 30_000,
242
+ pricing: {
243
+ units: [
244
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
245
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
246
+ ],
247
+ },
248
+ type: 'chat',
249
+ },
250
+ {
251
+ abilities: {
252
+ functionCall: true,
253
+ vision: true,
254
+ },
255
+ contextWindowTokens: 2_000_000,
256
+ description:
257
+ 'Grok 4 Fast（Non-Reasoning）是 xAI 的高吞吐、低成本多模态模型（支持 2M 上下文窗口），面向对延迟和成本敏感但不需要启用模型内推理的场景。它与 Grok 4 Fast 的 reasoning 版本并列，可通过 API 的 reasoning enable 参数在需要时开启推理功能。Prompts 和 completions 可能会被 xAI 或 OpenRouter 用于改进未来模型。',
258
+ displayName: 'Grok 4 Fast (Non-Reasoning)',
259
+ id: 'x-ai/grok-4-fast-non-reasoning',
260
+ maxOutput: 30_000,
261
+ pricing: {
262
+ units: [
263
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
264
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
265
+ ],
266
+ },
267
+ type: 'chat',
268
+ },
269
+ {
270
+ abilities: {
271
+ reasoning: true,
272
+ },
273
+ contextWindowTokens: 256_000,
274
+ description: 'Grok 4 是 xAI 的旗舰推理模型，提供强大的推理与多模态能力。',
275
+ displayName: 'Grok 4',
276
+ id: 'x-ai/grok-4',
277
+ maxOutput: 256_000,
278
+ pricing: {
279
+ units: [
280
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
281
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
282
+ ],
283
+ },
284
+ type: 'chat',
285
+ },
286
+ {
287
+ abilities: {
288
+ reasoning: true,
289
+ },
290
+ contextWindowTokens: 256_000,
291
+ description: 'Grok Code Fast 1 是 xAI 的快速代码模型，输出具可读性与工程化适配。',
292
+ displayName: 'Grok Code Fast 1',
293
+ id: 'x-ai/grok-code-fast-1',
294
+ maxOutput: 10_000,
295
+ pricing: {
296
+ units: [
297
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
298
+ { name: 'textOutput', rate: 1.5, strategy: 'fixed', unit: 'millionTokens' },
299
+ ],
300
+ },
301
+ type: 'chat',
302
+ },
303
+ {
304
+ abilities: {
305
+ reasoning: true,
306
+ vision: true,
307
+ },
308
+ contextWindowTokens: 128_000,
309
+ description:
310
+ 'ERNIE 5.0 Thinking Preview 是百度新一代原生多模态文心模型，擅长多模态理解、指令遵循、创作、事实问答与工具调用。',
311
+ displayName: 'ERNIE-5.0-Thinking-Preview',
312
+ id: 'baidu/ernie-5.0-thinking-preview',
313
+ maxOutput: 64_000,
314
+ pricing: {
315
+ units: [
316
+ { name: 'textInput', rate: 0.84, strategy: 'fixed', unit: 'millionTokens' },
317
+ { name: 'textOutput', rate: 3.37, strategy: 'fixed', unit: 'millionTokens' },
318
+ ],
319
+ },
320
+ type: 'chat',
321
+ },
322
+ {
323
+ abilities: {
324
+ functionCall: true,
325
+ reasoning: true,
326
+ vision: true,
327
+ },
328
+ contextWindowTokens: 256_000,
329
+ description:
330
+ 'Doubao-Seed-Code 是字节火山引擎面向 Agentic Programming 优化的大模型，在多项编程与代理基准上表现优异，支持 256K 上下文。',
331
+ displayName: 'Doubao-Seed-Code',
332
+ id: 'volcengine/doubao-seed-code',
333
+ maxOutput: 32_000,
334
+ pricing: {
335
+ units: [
336
+ { name: 'textInput', rate: 0.17, strategy: 'fixed', unit: 'millionTokens' },
337
+ { name: 'textOutput', rate: 1.12, strategy: 'fixed', unit: 'millionTokens' },
338
+ ],
339
+ },
340
+ type: 'chat',
341
+ },
342
+ {
343
+ abilities: {
344
+ functionCall: true,
345
+ reasoning: true,
346
+ vision: true,
347
+ },
348
+ contextWindowTokens: 262_144,
349
+ description:
350
+ 'Kimi K2 Thinking 是 Moonshot 针对深度推理任务优化的思考模型，具备通用 Agent 能力。',
351
+ displayName: 'Kimi K2 Thinking',
352
+ id: 'moonshotai/kimi-k2-thinking',
353
+ maxOutput: 262_144,
354
+ pricing: {
355
+ units: [
356
+ { name: 'textInput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
357
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
358
+ ],
359
+ },
360
+ type: 'chat',
361
+ },
362
+ {
363
+ abilities: {
364
+ functionCall: true,
365
+ reasoning: true,
366
+ vision: true,
367
+ },
368
+ contextWindowTokens: 262_144,
369
+ description:
370
+ 'Kimi K2 Thinking Turbo 是 Kimi K2 Thinking 的高速版本，在保持深度推理能力的同时，显著降低响应延迟。',
371
+ displayName: 'Kimi K2 Thinking Turbo',
372
+ id: 'moonshotai/kimi-k2-thinking-turbo',
373
+ maxOutput: 262_144,
374
+ pricing: {
375
+ units: [
376
+ { name: 'textInput', rate: 1.15, strategy: 'fixed', unit: 'millionTokens' },
377
+ { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
378
+ ],
379
+ },
380
+ type: 'chat',
381
+ },
382
+ {
383
+ abilities: {
384
+ functionCall: true,
385
+ reasoning: true,
386
+ vision: true,
387
+ },
388
+ contextWindowTokens: 65_536,
389
+ description:
390
+ 'Ming-flash-omni Preview 是 inclusionAI 的多模态模型，支持语音、图像和视频输入，优化了图像渲染与语音识别能力。',
391
+ displayName: 'Ming-flash-omini Preview',
392
+ id: 'inclusionai/ming-flash-omini-preview',
393
+ maxOutput: 32_000,
394
+ pricing: {
395
+ units: [
396
+ { name: 'textInput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
397
+ { name: 'textOutput', rate: 1.8, strategy: 'fixed', unit: 'millionTokens' },
398
+ ],
399
+ },
400
+ type: 'chat',
401
+ },
402
+ {
403
+ abilities: {
404
+ reasoning: true,
405
+ },
406
+ contextWindowTokens: 128_000,
407
+ description:
408
+ 'Ring-1T 是 inclusionAI 的 trillion-parameter MoE 思考模型，适合大规模推理与研究类任务。',
409
+ displayName: 'Ring-1T',
410
+ id: 'inclusionai/ring-1t',
411
+ maxOutput: 32_000,
412
+ pricing: {
413
+ units: [
414
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
415
+ { name: 'textOutput', rate: 2.24, strategy: 'fixed', unit: 'millionTokens' },
416
+ ],
417
+ },
418
+ type: 'chat',
419
+ },
420
+ {
421
+ abilities: {
422
+ reasoning: true,
423
+ },
424
+ contextWindowTokens: 128_000,
425
+ description:
426
+ 'Ling-1T 是 inclusionAI 的 1T MoE 大模型，针对高强度推理任务与大规模上下文进行了优化。',
427
+ displayName: 'Ling-1T',
428
+ id: 'inclusionai/ling-1t',
429
+ maxOutput: 32_000,
430
+ pricing: {
431
+ units: [
432
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
433
+ { name: 'textOutput', rate: 2.24, strategy: 'fixed', unit: 'millionTokens' },
434
+ ],
435
+ },
436
+ type: 'chat',
437
+ },
438
+ {
439
+ abilities: {
440
+ reasoning: true,
441
+ },
442
+ contextWindowTokens: 128_000,
443
+ description:
444
+ 'Ling-flash-2.0 是 inclusionAI 的 MoE 模型，优化了效率与推理表现，适合中大型任务。',
445
+ displayName: 'Ling-flash-2.0',
446
+ id: 'inclusionai/ling-flash-2.0',
447
+ maxOutput: 32_000,
448
+ pricing: {
449
+ units: [
450
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
451
+ { name: 'textOutput', rate: 2.8, strategy: 'fixed', unit: 'millionTokens' },
452
+ ],
453
+ },
454
+ type: 'chat',
455
+ },
456
+ {
457
+ abilities: {
458
+ reasoning: true,
459
+ },
460
+ contextWindowTokens: 128_000,
461
+ description:
462
+ 'Ring-flash-2.0 是 inclusionAI 面向高吞吐场景的 Ring 模型变体，强调速度与成本效率。',
463
+ displayName: 'Ring-flash-2.0',
464
+ id: 'inclusionai/ring-flash-2.0',
465
+ maxOutput: 32_000,
466
+ pricing: {
467
+ units: [
468
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
469
+ { name: 'textOutput', rate: 2.8, strategy: 'fixed', unit: 'millionTokens' },
470
+ ],
471
+ },
472
+ type: 'chat',
473
+ },
474
+ {
475
+ abilities: {
476
+ reasoning: true,
477
+ },
478
+ contextWindowTokens: 128_000,
479
+ description:
480
+ 'Ling-mini-2.0 是 inclusionAI 的轻量化 MoE 模型，在保持推理能力的同时显著降低成本。',
481
+ displayName: 'Ling-mini-2.0',
482
+ id: 'inclusionai/ling-mini-2.0',
483
+ maxOutput: 32_000,
484
+ pricing: {
485
+ units: [
486
+ { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
487
+ { name: 'textOutput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
488
+ ],
489
+ },
490
+ type: 'chat',
491
+ },
492
+ {
493
+ abilities: {
494
+ reasoning: true,
495
+ },
496
+ contextWindowTokens: 128_000,
497
+ description: 'Ring-mini-2.0 是 inclusionAI 的高吞吐轻量化 MoE 版本，主要用于并发场景。',
498
+ displayName: 'Ring-mini-2.0',
499
+ id: 'inclusionai/ring-mini-2.0',
500
+ maxOutput: 32_000,
501
+ pricing: {
502
+ units: [
503
+ { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
504
+ { name: 'textOutput', rate: 0.7, strategy: 'fixed', unit: 'millionTokens' },
505
+ ],
506
+ },
507
+ type: 'chat',
508
+ },
509
+ {
510
+ abilities: {
511
+ reasoning: true,
512
+ },
513
+ contextWindowTokens: 204_800,
514
+ description: 'MiniMax-M2 是一款在编码与代理任务上表现出色的高性价比模型，适合多种工程场景。',
515
+ displayName: 'MiniMax M2',
516
+ id: 'minimax/minimax-m2',
517
+ maxOutput: 128_000,
518
+ pricing: {
519
+ units: [
520
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
521
+ { name: 'textOutput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
522
+ ],
523
+ },
524
+ type: 'chat',
525
+ },
526
+ {
527
+ contextWindowTokens: 256_000,
528
+ description: 'KAT-Coder-Pro-V1（限时免费）专注于代码理解与自动化编程，用于高效的编程代理任务。',
529
+ displayName: 'KAT-Coder-Pro-V1 (Limited-time Free)',
530
+ id: 'kuaishou/kat-coder-pro-v1',
531
+ maxOutput: 32_000,
532
+ pricing: {
533
+ units: [
534
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
535
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
536
+ ],
537
+ },
538
+ type: 'chat',
539
+ },
540
+ {
541
+ abilities: {
542
+ functionCall: true,
543
+ reasoning: true,
544
+ search: true,
545
+ vision: true,
546
+ },
547
+ contextWindowTokens: 200_000,
548
+ description:
549
+ 'Claude Haiku 4.5 是 Anthropic 的高性能快速模型，在保持较高准确性的同时具有极低延迟。',
550
+ displayName: 'Claude Haiku 4.5',
551
+ id: 'anthropic/claude-haiku-4.5',
552
+ maxOutput: 64_000,
553
+ pricing: {
554
+ units: [
555
+ { name: 'textInput', rate: 1, strategy: 'fixed', unit: 'millionTokens' },
556
+ { name: 'textOutput', rate: 5, strategy: 'fixed', unit: 'millionTokens' },
557
+ ],
558
+ },
559
+ settings: {
560
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
561
+ searchImpl: 'params',
562
+ },
563
+ type: 'chat',
564
+ },
565
+ {
566
+ abilities: {
567
+ reasoning: true,
568
+ },
569
+ contextWindowTokens: 128_000,
570
+ description: 'DeepSeek-V3 是 DeepSeek 团队的一款高性能混合推理模型，适合复杂任务与工具集成。',
571
+ displayName: 'DeepSeek-V3.2-Exp (Non-thinking Mode)',
572
+ id: 'deepseek/deepseek-chat',
573
+ maxOutput: 8000,
574
+ pricing: {
575
+ units: [
576
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
577
+ { name: 'textOutput', rate: 1.68, strategy: 'fixed', unit: 'millionTokens' },
578
+ ],
579
+ },
580
+ type: 'chat',
581
+ },
582
+ {
583
+ abilities: {
584
+ functionCall: true,
585
+ reasoning: true,
586
+ },
587
+ contextWindowTokens: 128_000,
588
+ description:
589
+ 'DeepSeek-V3 Thinking（reasoner）是 DeepSeek 的实验 reasoning 模型，适合高复杂度推理任务。',
590
+ displayName: 'DeepSeek-V3.2-Exp (Thinking Mode)',
591
+ id: 'deepseek/deepseek-reasoner',
592
+ maxOutput: 64_000,
593
+ pricing: {
594
+ units: [
595
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
596
+ { name: 'textOutput', rate: 0.42, strategy: 'fixed', unit: 'millionTokens' },
597
+ ],
598
+ },
599
+ type: 'chat',
600
+ },
601
+ {
602
+ abilities: {
603
+ reasoning: true,
604
+ },
605
+ contextWindowTokens: 64_000,
606
+ description: 'DeepSeek R1 0528 是 DeepSeek 的更新变体，注重开源可用与推理深度。',
607
+ displayName: 'DeepSeek R1 0528',
608
+ id: 'deepseek/deepseek-r1-0528',
609
+ maxOutput: 64_000,
610
+ pricing: {
611
+ units: [
612
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
613
+ { name: 'textOutput', rate: 2.23, strategy: 'fixed', unit: 'millionTokens' },
614
+ ],
615
+ },
616
+ type: 'chat',
617
+ },
618
+ {
619
+ abilities: {
620
+ functionCall: true,
621
+ reasoning: true,
622
+ search: true,
623
+ vision: true,
624
+ },
625
+ contextWindowTokens: 1_050_000,
626
+ description: 'Gemini 2.5 Pro 是 Google 的旗舰级推理模型，支持长上下文与复杂任务。',
627
+ displayName: 'Gemini 2.5 Pro',
628
+ id: 'google/gemini-2.5-pro',
629
+ maxOutput: 65_530,
630
+ pricing: {
631
+ units: [
632
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
633
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
634
+ ],
635
+ },
636
+ settings: {
637
+ extendParams: ['thinkingBudget', 'urlContext'],
638
+ searchImpl: 'params',
639
+ searchProvider: 'google',
640
+ },
641
+ type: 'chat',
642
+ },
643
+ {
644
+ abilities: {
645
+ functionCall: true,
646
+ reasoning: true,
647
+ search: true,
648
+ vision: true,
649
+ },
650
+ contextWindowTokens: 1_050_000,
651
+ description: 'Gemini 2.5 Pro 免费版，支持受限额度的多模态长上下文，适合试用与轻量工作流。',
652
+ displayName: 'Gemini 2.5 Pro Free',
653
+ id: 'google/gemini-2.5-pro-free',
654
+ maxOutput: 65_530,
655
+ pricing: {
656
+ units: [
657
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
658
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
659
+ ],
660
+ },
661
+ settings: {
662
+ extendParams: ['thinkingBudget', 'urlContext'],
663
+ searchImpl: 'params',
664
+ searchProvider: 'google',
665
+ },
666
+ type: 'chat',
667
+ },
668
+ {
669
+ abilities: {
670
+ functionCall: true,
671
+ reasoning: true,
672
+ search: true,
673
+ vision: true,
674
+ },
675
+ contextWindowTokens: 1_050_000,
676
+ description: 'Gemini 2.5 Flash（Lite/Pro/Flash）系列是 Google 的中低延迟到高性能推理模型。',
677
+ displayName: 'Gemini 2.5 Flash',
678
+ id: 'google/gemini-2.5-flash',
679
+ maxOutput: 65_530,
680
+ pricing: {
681
+ units: [
682
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
683
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
684
+ ],
685
+ },
686
+ settings: {
687
+ extendParams: ['thinkingBudget', 'urlContext'],
688
+ searchImpl: 'params',
689
+ searchProvider: 'google',
690
+ },
691
+ type: 'chat',
692
+ },
693
+ {
694
+ abilities: {
695
+ functionCall: true,
696
+ reasoning: true,
697
+ vision: true,
698
+ },
699
+ contextWindowTokens: 400_000,
700
+ description:
701
+ 'GPT-5 Pro 是 OpenAI 的旗舰模型，提供更强的推理、代码生成与企业级功能，支持测试时路由与更严谨的安全策略。',
702
+ displayName: 'GPT-5 Pro',
703
+ id: 'openai/gpt-5-pro',
704
+ maxOutput: 128_000,
705
+ pricing: {
706
+ units: [
707
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
708
+ { name: 'textOutput', rate: 120, strategy: 'fixed', unit: 'millionTokens' },
709
+ ],
710
+ },
711
+ settings: {
712
+ extendParams: ['textVerbosity'],
713
+ },
714
+ type: 'chat',
715
+ },
716
+ {
717
+ abilities: {
718
+ functionCall: true,
719
+ reasoning: true,
720
+ vision: true,
721
+ },
722
+ contextWindowTokens: 400_000,
723
+ description: 'GPT-5 是 OpenAI 的高性能模型，适用广泛的生产与研究任务。',
724
+ displayName: 'GPT-5',
725
+ id: 'openai/gpt-5',
726
+ maxOutput: 128_000,
727
+ pricing: {
728
+ units: [
729
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
730
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
731
+ ],
732
+ },
733
+ settings: {
734
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
735
+ },
736
+ type: 'chat',
737
+ },
738
+ {
739
+ abilities: {
740
+ functionCall: true,
741
+ reasoning: true,
742
+ },
743
+ contextWindowTokens: 128_000,
744
+ description: 'GPT-5 Chat 是为对话场景优化的 GPT-5 子型号，降低延迟以提升交互体验。',
745
+ displayName: 'GPT-5 Chat',
746
+ id: 'openai/gpt-5-chat',
747
+ maxOutput: 16_380,
748
+ pricing: {
749
+ units: [
750
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
751
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
752
+ ],
753
+ },
754
+ type: 'chat',
755
+ },
756
+ {
757
+ abilities: {
758
+ functionCall: true,
759
+ reasoning: true,
760
+ vision: true,
761
+ },
762
+ contextWindowTokens: 400_000,
763
+ description: 'GPT-5 Mini 是 GPT-5 家族的精简版，适用于低延迟低成本场景。',
764
+ displayName: 'GPT-5 Mini',
765
+ id: 'openai/gpt-5-mini',
766
+ maxOutput: 128_000,
767
+ pricing: {
768
+ units: [
769
+ { name: 'textInput', rate: 0.25, strategy: 'fixed', unit: 'millionTokens' },
770
+ { name: 'textOutput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
771
+ ],
772
+ },
773
+ settings: {
774
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
775
+ },
776
+ type: 'chat',
777
+ },
778
+ {
779
+ abilities: {
780
+ functionCall: true,
781
+ reasoning: true,
782
+ },
783
+ contextWindowTokens: 400_000,
784
+ description: 'GPT-5 Nano 是家族中的超小型版本，适合对成本和延迟要求非常高的场景。',
785
+ displayName: 'GPT-5 Nano',
786
+ id: 'openai/gpt-5-nano',
787
+ maxOutput: 128_000,
788
+ pricing: {
789
+ units: [
790
+ { name: 'textInput', rate: 0.05, strategy: 'fixed', unit: 'millionTokens' },
791
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
792
+ ],
793
+ },
794
+ settings: {
795
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
796
+ },
797
+ type: 'chat',
798
+ },
799
+ {
800
+ abilities: {
801
+ functionCall: true,
802
+ reasoning: true,
803
+ },
804
+ contextWindowTokens: 400_000,
805
+ description: 'GPT-5-Codex 是针对编码场景进一步优化的 GPT-5 变体，适合大规模代码工作流。',
806
+ displayName: 'GPT-5 Codex',
807
+ id: 'openai/gpt-5-codex',
808
+ maxOutput: 128_000,
809
+ pricing: {
810
+ units: [
811
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
812
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
813
+ ],
814
+ },
815
+ settings: {
816
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
817
+ },
818
+ type: 'chat',
819
+ },
820
+ {
821
+ abilities: {
822
+ functionCall: true,
823
+ reasoning: true,
824
+ },
825
+ contextWindowTokens: 1_050_000,
826
+ description: 'GPT-4.1 系列提供了更大上下文与更强的工程与推理能力。',
827
+ displayName: 'GPT-4.1',
828
+ id: 'openai/gpt-4.1',
829
+ maxOutput: 32_770,
830
+ pricing: {
831
+ units: [
832
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
833
+ { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
834
+ ],
835
+ },
836
+ type: 'chat',
837
+ },
838
+ {
839
+ abilities: {
840
+ reasoning: true,
841
+ },
842
+ contextWindowTokens: 1_050_000,
843
+ description: 'GPT-4.1 Mini 提供更低延迟与更佳性价比，适合中等上下文上下线路。',
844
+ displayName: 'GPT-4.1 Mini',
845
+ id: 'openai/gpt-4.1-mini',
846
+ maxOutput: 32_770,
847
+ pricing: {
848
+ units: [
849
+ { name: 'textInput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
850
+ { name: 'textOutput', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
851
+ ],
852
+ },
853
+ type: 'chat',
854
+ },
855
+ {
856
+ abilities: {
857
+ reasoning: true,
858
+ },
859
+ contextWindowTokens: 1_050_000,
860
+ description: 'GPT-4.1 Nano 是极低成本低延迟选项，适合高频次短对话或分类场景。',
861
+ displayName: 'GPT-4.1 Nano',
862
+ id: 'openai/gpt-4.1-nano',
863
+ maxOutput: 32_770,
864
+ pricing: {
865
+ units: [
866
+ { name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
867
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
868
+ ],
869
+ },
870
+ type: 'chat',
871
+ },
872
+ {
873
+ abilities: {
874
+ reasoning: true,
875
+ },
876
+ contextWindowTokens: 262_144,
877
+ description: 'Qwen3 Max (preview) 是 Qwen 系列面向高级推理与工具集成的 Max 版本（预览）。',
878
+ displayName: 'Qwen3 Max Thinking Preview',
879
+ id: 'qwen/qwen3-max-preview',
880
+ maxOutput: 65_540,
881
+ pricing: {
882
+ units: [
883
+ { name: 'textInput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
884
+ { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
885
+ ],
886
+ },
887
+ type: 'chat',
888
+ },
889
+ {
890
+ abilities: {
891
+ reasoning: true,
892
+ },
893
+ contextWindowTokens: 262_144,
894
+ description: 'Qwen3 Max 是 Qwen3 系列的高端推理模型，适合多语言推理和工具集成。',
895
+ displayName: 'Qwen3 Max',
896
+ id: 'qwen/qwen3-max',
897
+ maxOutput: 65_540,
898
+ pricing: {
899
+ units: [
900
+ { name: 'textInput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
901
+ { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
902
+ ],
903
+ },
904
+ type: 'chat',
905
+ },
906
+ {
907
+ abilities: {
908
+ reasoning: true,
909
+ vision: true,
910
+ },
911
+ contextWindowTokens: 262_144,
912
+ description: 'Qwen3 VL-Plus 为 Qwen3 的视觉增强版本，提升了多模态推理与视频处理的能力。',
913
+ displayName: 'Qwen3-VL-Plus',
914
+ id: 'qwen/qwen3-vl-plus',
915
+ maxOutput: 32_770,
916
+ pricing: {
917
+ units: [
918
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
919
+ { name: 'textOutput', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
920
+ ],
921
+ },
922
+ type: 'chat',
923
+ },
924
+ {
925
+ abilities: {
926
+ reasoning: true,
927
+ },
928
+ contextWindowTokens: 1_000_000,
929
+ description:
930
+ 'Qwen3-Coder-Plus 为 Qwen 系列特别优化的编码代理模型，支持更复杂的工具调用与长期会话。',
931
+ displayName: 'Qwen3-Coder-Plus',
932
+ id: 'qwen/qwen3-coder-plus',
933
+ maxOutput: 65_540,
934
+ pricing: {
935
+ units: [
936
+ { name: 'textInput', rate: 1, strategy: 'fixed', unit: 'millionTokens' },
937
+ { name: 'textOutput', rate: 5, strategy: 'fixed', unit: 'millionTokens' },
938
+ ],
939
+ },
940
+ type: 'chat',
941
+ },
942
+ {
943
+ abilities: {
944
+ functionCall: true,
945
+ reasoning: true,
946
+ },
947
+ contextWindowTokens: 256_000,
948
+ description: 'Qwen3-Coder 是 Qwen3 的代码生成器家族，擅长长文档内的代码理解与生成。',
949
+ displayName: 'Qwen3-Coder',
950
+ id: 'qwen/qwen3-coder',
951
+ maxOutput: 261_000,
952
+ pricing: {
953
+ units: [
954
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
955
+ { name: 'textOutput', rate: 5.01, strategy: 'fixed', unit: 'millionTokens' },
956
+ ],
957
+ },
958
+ type: 'chat',
959
+ },
960
+ {
961
+ abilities: {
962
+ reasoning: true,
963
+ },
964
+ contextWindowTokens: 32_000,
965
+ description: 'Qwen3-14B 是 Qwen 系列的 14B 版本，适合常规推理与对话场景。',
966
+ displayName: 'Qwen3 14B',
967
+ id: 'qwen/qwen3-14b',
968
+ maxOutput: 32_000,
969
+ pricing: {
970
+ units: [
971
+ { name: 'textInput', rate: 0.14, strategy: 'fixed', unit: 'millionTokens' },
972
+ { name: 'textOutput', rate: 1.4, strategy: 'fixed', unit: 'millionTokens' },
973
+ ],
974
+ },
975
+ type: 'chat',
976
+ },
977
+ {
978
+ abilities: {
979
+ reasoning: true,
980
+ },
981
+ contextWindowTokens: 200_000,
982
+ description: 'GLM 4.6 是 Z.AI 的旗舰模型，扩展了上下文长度和编码能力。',
983
+ displayName: 'GLM 4.6',
984
+ id: 'z-ai/glm-4.6',
985
+ maxOutput: 128_000,
986
+ pricing: {
987
+ units: [
988
+ { name: 'textInput', rate: 0.35, strategy: 'fixed', unit: 'millionTokens' },
989
+ { name: 'textOutput', rate: 1.54, strategy: 'fixed', unit: 'millionTokens' },
990
+ ],
991
+ },
992
+ type: 'chat',
993
+ },
994
+ {
995
+ abilities: {
996
+ functionCall: true,
997
+ reasoning: true,
998
+ search: true,
999
+ vision: true,
1000
+ },
1001
+ contextWindowTokens: 200_000,
1002
+ description: 'Opus 4.1 是 Anthropic 的高端模型，优化于编程、复杂推理和持续任务。',
1003
+ displayName: 'Claude Opus 4.1',
1004
+ id: 'anthropic/claude-opus-4.1',
1005
+ maxOutput: 32_000,
1006
+ pricing: {
1007
+ units: [
1008
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1009
+ { name: 'textOutput', rate: 75, strategy: 'fixed', unit: 'millionTokens' },
1010
+ ],
1011
+ },
1012
+ settings: {
1013
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1014
+ searchImpl: 'params',
1015
+ },
1016
+ type: 'chat',
1017
+ },
1018
+ {
1019
+ abilities: {
1020
+ functionCall: true,
1021
+ reasoning: true,
1022
+ search: true,
1023
+ vision: true,
1024
+ },
1025
+ contextWindowTokens: 200_000,
1026
+ description: 'Opus 4 是 Anthropic 的旗舰级模型，专为复杂任务和企业级应用设计。',
1027
+ displayName: 'Claude Opus 4',
1028
+ id: 'anthropic/claude-opus-4',
1029
+ maxOutput: 32_000,
1030
+ pricing: {
1031
+ units: [
1032
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1033
+ { name: 'textOutput', rate: 75, strategy: 'fixed', unit: 'millionTokens' },
1034
+ ],
1035
+ },
1036
+ settings: {
1037
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1038
+ searchImpl: 'params',
1039
+ },
1040
+ type: 'chat',
1041
+ },
1042
+ {
1043
+ abilities: {
1044
+ functionCall: true,
1045
+ imageOutput: true,
1046
+ reasoning: true,
1047
+ vision: true,
1048
+ },
1049
+ contextWindowTokens: 32_768,
1050
+ description:
1051
+ 'Gemini 2.5 Flash Image（Nano Banana）是 Google 的图像生成模型，同时支持多模态对话。',
1052
+ displayName: 'Gemini 2.5 Flash Image (Nano Banana)',
1053
+ id: 'google/gemini-2.5-flash-image',
1054
+ maxOutput: 8192,
1055
+ pricing: {
1056
+ approximatePricePerImage: 0.039,
1057
+ units: [
1058
+ { name: 'imageOutput', rate: 30, strategy: 'fixed', unit: 'millionTokens' },
1059
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
1060
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1061
+ ],
1062
+ },
1063
+ type: 'chat',
1064
+ },
1065
+ {
1066
+ abilities: {
1067
+ functionCall: true,
1068
+ imageOutput: true,
1069
+ reasoning: true,
1070
+ vision: true,
1071
+ },
1072
+ contextWindowTokens: 32_768,
1073
+ description: 'Gemini 2.5 Flash Image 免费版，支持受限额度的多模态生成。',
1074
+ displayName: 'Gemini 2.5 Flash Image (Nano Banana) Free',
1075
+ id: 'google/gemini-2.5-flash-image-free',
1076
+ maxOutput: 8192,
1077
+ pricing: {
1078
+ units: [
1079
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
1080
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
1081
+ ],
1082
+ },
1083
+ type: 'chat',
1084
+ },
1085
+ {
1086
+ abilities: {
1087
+ functionCall: true,
1088
+ reasoning: true,
1089
+ search: true,
1090
+ vision: true,
1091
+ },
1092
+ contextWindowTokens: 200_000,
1093
+ description: 'Claude Sonnet 4.5 是 Anthropic 最新的混合推理模型，优化于复杂推理和编码。',
1094
+ displayName: 'Claude Sonnet 4.5',
1095
+ id: 'anthropic/claude-sonnet-4.5',
1096
+ maxOutput: 64_000,
1097
+ pricing: {
1098
+ units: [
1099
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1100
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1101
+ ],
1102
+ },
1103
+ settings: {
1104
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1105
+ searchImpl: 'params',
1106
+ },
1107
+ type: 'chat',
1108
+ },
1109
+ {
1110
+ abilities: {
1111
+ functionCall: true,
1112
+ reasoning: true,
1113
+ },
1114
+ contextWindowTokens: 1_000_000,
1115
+ description: 'Claude Sonnet 4 是 Anthropic 的混合推理版本，提供思维/非思维混合能力。',
1116
+ displayName: 'Claude Sonnet 4',
1117
+ id: 'anthropic/claude-sonnet-4',
1118
+ maxOutput: 64_000,
1119
+ pricing: {
1120
+ units: [
1121
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1122
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1123
+ ],
1124
+ },
1125
+ type: 'chat',
1126
+ },
1127
+ {
1128
+ abilities: {
1129
+ functionCall: true,
1130
+ reasoning: true,
1131
+ },
1132
+ contextWindowTokens: 200_000,
1133
+ description: 'OpenAI o4-mini 是 OpenAI 的小型高效推理模型，适合低延迟场景。',
1134
+ displayName: 'o4 Mini',
1135
+ id: 'openai/o4-mini',
1136
+ maxOutput: 100_000,
1137
+ pricing: {
1138
+ units: [
1139
+ { name: 'textInput', rate: 1.1, strategy: 'fixed', unit: 'millionTokens' },
1140
+ { name: 'textOutput', rate: 4.4, strategy: 'fixed', unit: 'millionTokens' },
1141
+ ],
1142
+ },
1143
+ settings: {
1144
+ extendParams: ['reasoningEffort'],
1145
+ },
1146
+ type: 'chat',
1147
+ },
1148
+ {
1149
+ abilities: {
1150
+ functionCall: true,
1151
+ reasoning: true,
1152
+ },
1153
+ contextWindowTokens: 128_000,
1154
+ description: 'GPT-4o 系列是 OpenAI 的 Omni 模型，支持文本 + 图片输入与文本输出。',
1155
+ displayName: 'GPT-4o',
1156
+ id: 'openai/gpt-4o',
1157
+ maxOutput: 16_380,
1158
+ pricing: {
1159
+ units: [
1160
+ { name: 'textInput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1161
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
1162
+ ],
1163
+ },
1164
+ type: 'chat',
1165
+ },
1166
+ {
1167
+ abilities: {
1168
+ functionCall: true,
1169
+ reasoning: true,
1170
+ },
1171
+ contextWindowTokens: 128_000,
1172
+ description: 'GPT-4o-mini 是 GPT-4o 的快速小模型版本，适合低延迟图文混合场景。',
1173
+ displayName: 'GPT-4o-mini',
1174
+ id: 'openai/gpt-4o-mini',
1175
+ maxOutput: 16_380,
1176
+ pricing: {
1177
+ units: [
1178
+ { name: 'textInput', rate: 0.15, strategy: 'fixed', unit: 'millionTokens' },
1179
+ { name: 'textOutput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1180
+ ],
1181
+ },
1182
+ type: 'chat',
1183
+ },
1184
+ {
1185
+ abilities: {
1186
+ reasoning: true,
1187
+ },
1188
+ contextWindowTokens: 128_000,
1189
+ description: 'Kimi K2 0711 是 Kimi 系列的 Instruct 版本，适合高质量代码与工具调用场景。',
1190
+ displayName: 'Kimi K2 0711',
1191
+ id: 'moonshotai/kimi-k2-0711',
1192
+ maxOutput: 32_000,
1193
+ pricing: {
1194
+ units: [
1195
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
1196
+ { name: 'textOutput', rate: 2.23, strategy: 'fixed', unit: 'millionTokens' },
1197
+ ],
1198
+ },
1199
+ type: 'chat',
1200
+ },
1201
+ {
1202
+ abilities: {
1203
+ reasoning: true,
1204
+ },
1205
+ contextWindowTokens: 128_000,
1206
+ description: 'GLM 4.5 Air 是 GLM 4.5 的轻量化版本，适合成本敏感场景但保留强推理能力。',
1207
+ displayName: 'GLM 4.5 Air',
1208
+ id: 'z-ai/glm-4.5-air',
1209
+ maxOutput: 96_000,
1210
+ pricing: {
1211
+ units: [
1212
+ { name: 'textInput', rate: 0.11, strategy: 'fixed', unit: 'millionTokens' },
1213
+ { name: 'textOutput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
1214
+ ],
1215
+ },
1216
+ type: 'chat',
1217
+ },
1218
+ {
1219
+ abilities: {
1220
+ functionCall: true,
1221
+ reasoning: true,
1222
+ },
1223
+ contextWindowTokens: 200_000,
1224
+ description:
1225
+ 'Claude 3.5 Haiku features enhanced capabilities in speed, coding accuracy, and tool use. 适用于对速度与工具交互有高要求的场景。',
1226
+ displayName: 'Claude 3.5 Haiku',
1227
+ id: 'anthropic/claude-3.5-haiku',
1228
+ maxOutput: 8192,
1229
+ pricing: {
1230
+ units: [
1231
+ { name: 'textInput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
1232
+ { name: 'textOutput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
1233
+ ],
1234
+ },
1235
+ type: 'chat',
1236
+ },
1237
+ {
1238
+ abilities: {
1239
+ functionCall: true,
1240
+ reasoning: true,
1241
+ },
1242
+ contextWindowTokens: 200_000,
1243
+ description:
1244
+ 'Claude 3.5 Sonnet 是 Sonnet 家族的快速高效模型，提供更好的编码与推理性能，部分版本将逐步被 Sonnet 3.7 等替代。',
1245
+ displayName: 'Claude 3.5 Sonnet',
1246
+ id: 'anthropic/claude-3.5-sonnet',
1247
+ maxOutput: 8192,
1248
+ pricing: {
1249
+ units: [
1250
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1251
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1252
+ ],
1253
+ },
1254
+ type: 'chat',
1255
+ },
1256
+ {
1257
+ abilities: {
1258
+ functionCall: true,
1259
+ reasoning: true,
1260
+ },
1261
+ contextWindowTokens: 200_000,
1262
+ description:
1263
+ 'Claude 3.7 Sonnet 是 Sonnet 系列的升级版，提供更强的推理与编码能力，适用于企业级复杂任务。',
1264
+ displayName: 'Claude 3.7 Sonnet',
1265
+ id: 'anthropic/claude-3.7-sonnet',
1266
+ maxOutput: 64_000,
1267
+ pricing: {
1268
+ units: [
1269
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1270
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1271
+ ],
1272
+ },
1273
+ type: 'chat',
1274
+ },
1275
+ {
1276
+ abilities: {
1277
+ functionCall: true,
1278
+ reasoning: true,
1279
+ search: true,
1280
+ },
1281
+ contextWindowTokens: 128_000,
1282
+ description:
1283
+ 'DeepSeek-V3.1 是 DeepSeek 的长上下文混合推理模型，支持思考/非思考混合模式与工具集成。',
1284
+ displayName: 'DeepSeek V3.1',
1285
+ id: 'deepseek/deepseek-chat-v3.1',
1286
+ maxOutput: 65_540,
1287
+ pricing: {
1288
+ units: [
1289
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1290
+ { name: 'textOutput', rate: 1.11, strategy: 'fixed', unit: 'millionTokens' },
1291
+ ],
1292
+ },
1293
+ type: 'chat',
1294
+ },
1295
+ {
1296
+ abilities: {
1297
+ functionCall: true,
1298
+ reasoning: true,
1299
+ vision: true,
1300
+ },
1301
+ contextWindowTokens: 1_050_000,
1302
+ description: 'Gemini 2.0 Flash 是 Google 的高性能推理模型，适用于延展的多模态任务。',
1303
+ displayName: 'Gemini 2.0 Flash',
1304
+ id: 'google/gemini-2.0-flash',
1305
+ maxOutput: 8192,
1306
+ pricing: {
1307
+ units: [
1308
+ { name: 'textInput', rate: 0.15, strategy: 'fixed', unit: 'millionTokens' },
1309
+ { name: 'textOutput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1310
+ ],
1311
+ },
1312
+ type: 'chat',
1313
+ },
1314
+ {
1315
+ abilities: {
1316
+ functionCall: true,
1317
+ reasoning: true,
1318
+ vision: true,
1319
+ },
1320
+ contextWindowTokens: 1_050_000,
1321
+ description:
1322
+ 'Gemini 2.0 Flash Lite 是 Gemini 家族的轻量版本，默认不启用思考以提升延迟与成本表现，但可通过参数开启。',
1323
+ displayName: 'Gemini 2.0 Flash Lite',
1324
+ id: 'google/gemini-2.0-flash-lite-001',
1325
+ maxOutput: 8192,
1326
+ pricing: {
1327
+ units: [
1328
+ { name: 'textInput', rate: 0.075, strategy: 'fixed', unit: 'millionTokens' },
1329
+ { name: 'textOutput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
1330
+ ],
1331
+ },
1332
+ type: 'chat',
1333
+ },
1334
+ {
1335
+ abilities: {
1336
+ functionCall: true,
1337
+ reasoning: true,
1338
+ vision: true,
1339
+ },
1340
+ contextWindowTokens: 1_050_000,
1341
+ description:
1342
+ 'Gemini 2.5 Flash Lite 是 Gemini 2.5 的轻量版本，优化了延迟与成本，适合高吞吐场景。',
1343
+ displayName: 'Gemini 2.5 Flash Lite',
1344
+ id: 'google/gemini-2.5-flash-lite',
1345
+ maxOutput: 65_530,
1346
+ pricing: {
1347
+ units: [
1348
+ { name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
1349
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
1350
+ ],
1351
+ },
1352
+ type: 'chat',
1353
+ },
1354
+ {
1355
+ abilities: {
1356
+ reasoning: true,
1357
+ },
1358
+ contextWindowTokens: 262_144,
1359
+ description: 'Kimi K2 0905 是 Kimi 系列的 0905 更新，扩充了上下文与推理性能，优化了编码场景。',
1360
+ displayName: 'Kimi K2 0905',
1361
+ id: 'moonshotai/kimi-k2-0905',
1362
+ maxOutput: 262_144,
1363
+ pricing: {
1364
+ units: [
1365
+ { name: 'textInput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1366
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1367
+ ],
1368
+ },
1369
+ type: 'chat',
1370
+ },
1371
+ {
1372
+ abilities: {
1373
+ reasoning: true,
1374
+ },
1375
+ contextWindowTokens: 256_000,
1376
+ description:
1377
+ 'Qwen3-235B-A22B-Instruct-2507 为 Qwen3 系列的 Instruct 版本，兼顾多语言指令与长上下文场景。',
1378
+ displayName: 'Qwen3 235B A22B Instruct 2507',
1379
+ id: 'qwen/qwen3-235b-a22b-2507',
1380
+ maxOutput: 262_100,
1381
+ pricing: {
1382
+ units: [
1383
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1384
+ { name: 'textOutput', rate: 1.11, strategy: 'fixed', unit: 'millionTokens' },
1385
+ ],
1386
+ },
1387
+ type: 'chat',
1388
+ },
1389
+ {
1390
+ abilities: {
1391
+ reasoning: true,
1392
+ },
1393
+ contextWindowTokens: 256_000,
1394
+ description:
1395
+ 'Qwen3-235B-A22B-Thinking-2507 为 Qwen3 的 Thinking 变体，针对复杂数学与推理任务进行了强化。',
1396
+ displayName: 'Qwen3 235B A22B Thinking 2507',
1397
+ id: 'qwen/qwen3-235b-a22b-thinking-2507',
1398
+ maxOutput: 262_100,
1399
+ pricing: {
1400
+ units: [
1401
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1402
+ { name: 'textOutput', rate: 2.78, strategy: 'fixed', unit: 'millionTokens' },
1403
+ ],
1404
+ },
1405
+ type: 'chat',
1406
+ },
1407
+ {
1408
+ abilities: { reasoning: true },
1409
+ contextWindowTokens: 128_000,
1410
+ description: 'GLM 4.5 是 Z.AI 的旗舰模型，支持混合推理模式并优化于工程与长上下文任务。',
1411
+ displayName: 'GLM 4.5',
1412
+ id: 'z-ai/glm-4.5',
1413
+ maxOutput: 96_000,
1414
+ pricing: {
1415
+ units: [
1416
+ { name: 'textInput', rate: 0.35, strategy: 'fixed', unit: 'millionTokens' },
1417
+ { name: 'textOutput', rate: 1.54, strategy: 'fixed', unit: 'millionTokens' },
1418
+ ],
1419
+ },
1420
+ type: 'chat',
1421
+ },
1422
+ ];
1423
+
1424
+ export const allModels = [...zenmuxChatModels];
1425
+
1426
+ export default allModels;