@lobehub/lobehub 2.0.0-next.100 → 2.0.0-next.102

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 (40) hide show
  1. package/AGENTS.md +1 -0
  2. package/CHANGELOG.md +50 -0
  3. package/CLAUDE.md +1 -0
  4. package/changelog/v1.json +18 -0
  5. package/package.json +1 -1
  6. package/packages/model-bank/package.json +1 -0
  7. package/packages/model-bank/src/aiModels/aihubmix.ts +27 -0
  8. package/packages/model-bank/src/aiModels/google.ts +69 -10
  9. package/packages/model-bank/src/aiModels/index.ts +3 -0
  10. package/packages/model-bank/src/aiModels/infiniai.ts +5 -22
  11. package/packages/model-bank/src/aiModels/ollamacloud.ts +12 -0
  12. package/packages/model-bank/src/aiModels/siliconcloud.ts +0 -61
  13. package/packages/model-bank/src/aiModels/vertexai.ts +88 -1
  14. package/packages/model-bank/src/aiModels/zenmux.ts +1423 -0
  15. package/packages/model-bank/src/const/modelProvider.ts +1 -0
  16. package/packages/model-bank/src/standard-parameters/index.ts +9 -0
  17. package/packages/model-runtime/src/core/RouterRuntime/createRuntime.ts +42 -18
  18. package/packages/model-runtime/src/core/openaiCompatibleFactory/index.test.ts +2 -2
  19. package/packages/model-runtime/src/core/streams/bedrock/claude.ts +17 -3
  20. package/packages/model-runtime/src/core/streams/google/index.ts +7 -2
  21. package/packages/model-runtime/src/core/streams/openai/__snapshots__/responsesStream.test.ts.snap +166 -166
  22. package/packages/model-runtime/src/index.ts +1 -1
  23. package/packages/model-runtime/src/providers/anthropic/index.ts +1 -38
  24. package/packages/model-runtime/src/providers/anthropic/resolveCacheTTL.ts +44 -0
  25. package/packages/model-runtime/src/providers/bedrock/index.test.ts +127 -11
  26. package/packages/model-runtime/src/providers/bedrock/index.ts +47 -13
  27. package/packages/model-runtime/src/providers/google/createImage.ts +1 -0
  28. package/packages/model-runtime/src/providers/google/index.ts +11 -1
  29. package/packages/model-runtime/src/providers/zenmux/index.test.ts +320 -0
  30. package/packages/model-runtime/src/providers/zenmux/index.ts +84 -0
  31. package/packages/model-runtime/src/runtimeMap.ts +2 -0
  32. package/packages/types/src/user/settings/keyVaults.ts +1 -0
  33. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/components/ResolutionSelect.tsx +88 -0
  34. package/src/app/[variants]/(main)/image/@menu/features/ConfigPanel/index.tsx +9 -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/chat/slices/aiChat/actions/__tests__/conversationLifecycle.test.ts +3 -0
  40. package/src/store/chat/slices/aiChat/actions/streamingExecutor.ts +11 -0
package/packages/model-bank/src/aiModels/zenmux.ts ADDED
@@ -0,0 +1,1423 @@
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
+ id: 'google/gemini-3-pro-image-preview',
23
+ maxOutput: 32_768,
24
+ pricing: {
25
+ units: [
26
+ { name: 'imageOutput', rate: 60, strategy: 'fixed', unit: 'millionTokens' },
27
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
28
+ { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
29
+ ],
30
+ },
31
+ type: 'chat',
32
+ },
33
+ {
34
+ abilities: {
35
+ functionCall: true,
36
+ imageOutput: true,
37
+ reasoning: true,
38
+ vision: true,
39
+ },
40
+ contextWindowTokens: 65_536,
41
+ description: 'Gemini 3 Pro Image 免费版，支持受限额度的多模态生成。',
42
+ displayName: 'Gemini 3 Pro Image (Nano Banana) Free',
43
+ id: 'google/gemini-3-pro-image-preview-free',
44
+ maxOutput: 32_768,
45
+ pricing: {
46
+ units: [
47
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
48
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
49
+ ],
50
+ },
51
+ type: 'chat',
52
+ },
53
+ {
54
+ abilities: {
55
+ functionCall: true,
56
+ reasoning: true,
57
+ vision: true,
58
+ },
59
+ contextWindowTokens: 1_050_000,
60
+ description:
61
+ '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。',
62
+ displayName: 'Gemini 3 Pro Preview',
63
+ id: 'google/gemini-3-pro-preview',
64
+ maxOutput: 65_530,
65
+ pricing: {
66
+ units: [
67
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
68
+ { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
69
+ ],
70
+ },
71
+ type: 'chat',
72
+ },
73
+ {
74
+ abilities: {
75
+ functionCall: true,
76
+ reasoning: true,
77
+ search: true,
78
+ vision: true,
79
+ },
80
+ contextWindowTokens: 1_050_000,
81
+ description:
82
+ '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。',
83
+ displayName: 'Gemini 3 Pro Preview Free',
84
+ id: 'google/gemini-3-pro-preview-free',
85
+ maxOutput: 65_530,
86
+ pricing: {
87
+ units: [
88
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
89
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
90
+ ],
91
+ },
92
+ settings: {
93
+ extendParams: ['thinkingLevel', 'urlContext'],
94
+ searchImpl: 'params',
95
+ searchProvider: 'google',
96
+ },
97
+ type: 'chat',
98
+ },
99
+ {
100
+ abilities: {
101
+ functionCall: true,
102
+ reasoning: true,
103
+ vision: true,
104
+ },
105
+ contextWindowTokens: 400_000,
106
+ description:
107
+ '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。',
108
+ displayName: 'GPT-5.1',
109
+ id: 'openai/gpt-5.1',
110
+ maxOutput: 128_000,
111
+ pricing: {
112
+ units: [
113
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
114
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
115
+ ],
116
+ },
117
+ settings: {
118
+ extendParams: ['gpt5_1ReasoningEffort', 'textVerbosity'],
119
+ },
120
+ type: 'chat',
121
+ },
122
+ {
123
+ abilities: {
124
+ functionCall: true,
125
+ reasoning: true,
126
+ vision: true,
127
+ },
128
+ contextWindowTokens: 128_000,
129
+ description:
130
+ '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。',
131
+ displayName: 'GPT-5.1 Chat',
132
+ id: 'openai/gpt-5.1-chat',
133
+ maxOutput: 16_380,
134
+ pricing: {
135
+ units: [
136
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
137
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
138
+ ],
139
+ },
140
+ type: 'chat',
141
+ },
142
+ {
143
+ abilities: {
144
+ functionCall: true,
145
+ reasoning: true,
146
+ vision: true,
147
+ },
148
+ contextWindowTokens: 400_000,
149
+ description:
150
+ '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。',
151
+ displayName: 'GPT-5.1-Codex',
152
+ id: 'openai/gpt-5.1-codex',
153
+ maxOutput: 128_000,
154
+ pricing: {
155
+ units: [
156
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
157
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
158
+ ],
159
+ },
160
+ settings: {
161
+ extendParams: ['gpt5_1ReasoningEffort'],
162
+ },
163
+ type: 'chat',
164
+ },
165
+ {
166
+ abilities: {
167
+ functionCall: true,
168
+ reasoning: true,
169
+ vision: true,
170
+ },
171
+ contextWindowTokens: 400_000,
172
+ description:
173
+ '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。',
174
+ displayName: 'GPT-5.1-Codex-Mini',
175
+ id: 'openai/gpt-5.1-codex-mini',
176
+ maxOutput: 100_000,
177
+ pricing: {
178
+ units: [
179
+ { name: 'textInput', rate: 0.25, strategy: 'fixed', unit: 'millionTokens' },
180
+ { name: 'textOutput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
181
+ ],
182
+ },
183
+ settings: {
184
+ extendParams: ['gpt5_1ReasoningEffort'],
185
+ },
186
+ type: 'chat',
187
+ },
188
+ {
189
+ abilities: {
190
+ functionCall: true,
191
+ reasoning: true,
192
+ vision: true,
193
+ },
194
+ contextWindowTokens: 2_000_000,
195
+ description:
196
+ '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。',
197
+ displayName: 'Grok 4.1 Fast',
198
+ id: 'x-ai/grok-4.1-fast',
199
+ maxOutput: 30_000,
200
+ pricing: {
201
+ units: [
202
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
203
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
204
+ ],
205
+ },
206
+ type: 'chat',
207
+ },
208
+ {
209
+ abilities: {
210
+ functionCall: true,
211
+ vision: true,
212
+ },
213
+ contextWindowTokens: 2_000_000,
214
+ description:
215
+ '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。',
216
+ displayName: 'Grok 4.1 Fast (Non-Reasoning)',
217
+ id: 'x-ai/grok-4.1-fast-non-reasoning',
218
+ maxOutput: 30_000,
219
+ pricing: {
220
+ units: [
221
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
222
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
223
+ ],
224
+ },
225
+ type: 'chat',
226
+ },
227
+ {
228
+ abilities: {
229
+ functionCall: true,
230
+ reasoning: true,
231
+ vision: true,
232
+ },
233
+ contextWindowTokens: 2_000_000,
234
+ description:
235
+ '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。',
236
+ displayName: 'Grok 4 Fast',
237
+ id: 'x-ai/grok-4-fast',
238
+ maxOutput: 30_000,
239
+ pricing: {
240
+ units: [
241
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
242
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
243
+ ],
244
+ },
245
+ type: 'chat',
246
+ },
247
+ {
248
+ abilities: {
249
+ functionCall: true,
250
+ vision: true,
251
+ },
252
+ contextWindowTokens: 2_000_000,
253
+ description:
254
+ '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。',
255
+ displayName: 'Grok 4 Fast (Non-Reasoning)',
256
+ id: 'x-ai/grok-4-fast-non-reasoning',
257
+ maxOutput: 30_000,
258
+ pricing: {
259
+ units: [
260
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
261
+ { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
262
+ ],
263
+ },
264
+ type: 'chat',
265
+ },
266
+ {
267
+ abilities: {
268
+ reasoning: true,
269
+ },
270
+ contextWindowTokens: 256_000,
271
+ 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。',
272
+ displayName: 'Grok 4',
273
+ id: 'x-ai/grok-4',
274
+ maxOutput: 256_000,
275
+ pricing: {
276
+ units: [
277
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
278
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
279
+ ],
280
+ },
281
+ type: 'chat',
282
+ },
283
+ {
284
+ abilities: {
285
+ reasoning: true,
286
+ },
287
+ contextWindowTokens: 256_000,
288
+ 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。',
289
+ displayName: 'Grok Code Fast 1',
290
+ id: 'x-ai/grok-code-fast-1',
291
+ maxOutput: 10_000,
292
+ pricing: {
293
+ units: [
294
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
295
+ { name: 'textOutput', rate: 1.5, strategy: 'fixed', unit: 'millionTokens' },
296
+ ],
297
+ },
298
+ type: 'chat',
299
+ },
300
+ {
301
+ abilities: {
302
+ reasoning: true,
303
+ vision: true,
304
+ },
305
+ contextWindowTokens: 128_000,
306
+ description:
307
+ '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。',
308
+ displayName: 'ERNIE-5.0-Thinking-Preview',
309
+ id: 'baidu/ernie-5.0-thinking-preview',
310
+ maxOutput: 64_000,
311
+ pricing: {
312
+ units: [
313
+ { name: 'textInput', rate: 0.84, strategy: 'fixed', unit: 'millionTokens' },
314
+ { name: 'textOutput', rate: 3.37, strategy: 'fixed', unit: 'millionTokens' },
315
+ ],
316
+ },
317
+ type: 'chat',
318
+ },
319
+ {
320
+ abilities: {
321
+ functionCall: true,
322
+ reasoning: true,
323
+ vision: true,
324
+ },
325
+ contextWindowTokens: 256_000,
326
+ description:
327
+ '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。',
328
+ displayName: 'Doubao-Seed-Code',
329
+ id: 'volcengine/doubao-seed-code',
330
+ maxOutput: 32_000,
331
+ pricing: {
332
+ units: [
333
+ { name: 'textInput', rate: 0.17, strategy: 'fixed', unit: 'millionTokens' },
334
+ { name: 'textOutput', rate: 1.12, strategy: 'fixed', unit: 'millionTokens' },
335
+ ],
336
+ },
337
+ type: 'chat',
338
+ },
339
+ {
340
+ abilities: {
341
+ functionCall: true,
342
+ reasoning: true,
343
+ vision: true,
344
+ },
345
+ contextWindowTokens: 262_144,
346
+ description:
347
+ '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。',
348
+ displayName: 'Kimi K2 Thinking',
349
+ id: 'moonshotai/kimi-k2-thinking',
350
+ maxOutput: 262_144,
351
+ pricing: {
352
+ units: [
353
+ { name: 'textInput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
354
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
355
+ ],
356
+ },
357
+ type: 'chat',
358
+ },
359
+ {
360
+ abilities: {
361
+ functionCall: true,
362
+ reasoning: true,
363
+ vision: true,
364
+ },
365
+ contextWindowTokens: 262_144,
366
+ description:
367
+ '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。',
368
+ displayName: 'Kimi K2 Thinking Turbo',
369
+ id: 'moonshotai/kimi-k2-thinking-turbo',
370
+ maxOutput: 262_144,
371
+ pricing: {
372
+ units: [
373
+ { name: 'textInput', rate: 1.15, strategy: 'fixed', unit: 'millionTokens' },
374
+ { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
375
+ ],
376
+ },
377
+ type: 'chat',
378
+ },
379
+ {
380
+ abilities: {
381
+ functionCall: true,
382
+ reasoning: true,
383
+ vision: true,
384
+ },
385
+ contextWindowTokens: 65_536,
386
+ description:
387
+ '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。',
388
+ displayName: 'Ming-flash-omini Preview',
389
+ id: 'inclusionai/ming-flash-omini-preview',
390
+ maxOutput: 32_000,
391
+ pricing: {
392
+ units: [
393
+ { name: 'textInput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
394
+ { name: 'textOutput', rate: 1.8, strategy: 'fixed', unit: 'millionTokens' },
395
+ ],
396
+ },
397
+ type: 'chat',
398
+ },
399
+ {
400
+ abilities: {
401
+ reasoning: true,
402
+ },
403
+ contextWindowTokens: 128_000,
404
+ description:
405
+ '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。',
406
+ displayName: 'Ring-1T',
407
+ id: 'inclusionai/ring-1t',
408
+ maxOutput: 32_000,
409
+ pricing: {
410
+ units: [
411
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
412
+ { name: 'textOutput', rate: 2.24, strategy: 'fixed', unit: 'millionTokens' },
413
+ ],
414
+ },
415
+ type: 'chat',
416
+ },
417
+ {
418
+ abilities: {
419
+ reasoning: true,
420
+ },
421
+ contextWindowTokens: 128_000,
422
+ description:
423
+ '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。',
424
+ displayName: 'Ling-1T',
425
+ id: 'inclusionai/ling-1t',
426
+ maxOutput: 32_000,
427
+ pricing: {
428
+ units: [
429
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
430
+ { name: 'textOutput', rate: 2.24, strategy: 'fixed', unit: 'millionTokens' },
431
+ ],
432
+ },
433
+ type: 'chat',
434
+ },
435
+ {
436
+ abilities: {
437
+ reasoning: true,
438
+ },
439
+ contextWindowTokens: 128_000,
440
+ description:
441
+ '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。',
442
+ displayName: 'Ling-flash-2.0',
443
+ id: 'inclusionai/ling-flash-2.0',
444
+ maxOutput: 32_000,
445
+ pricing: {
446
+ units: [
447
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
448
+ { name: 'textOutput', rate: 2.8, strategy: 'fixed', unit: 'millionTokens' },
449
+ ],
450
+ },
451
+ type: 'chat',
452
+ },
453
+ {
454
+ abilities: {
455
+ reasoning: true,
456
+ },
457
+ contextWindowTokens: 128_000,
458
+ description:
459
+ '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。',
460
+ displayName: 'Ring-flash-2.0',
461
+ id: 'inclusionai/ring-flash-2.0',
462
+ maxOutput: 32_000,
463
+ pricing: {
464
+ units: [
465
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
466
+ { name: 'textOutput', rate: 2.8, strategy: 'fixed', unit: 'millionTokens' },
467
+ ],
468
+ },
469
+ type: 'chat',
470
+ },
471
+ {
472
+ abilities: {
473
+ reasoning: true,
474
+ },
475
+ contextWindowTokens: 128_000,
476
+ description:
477
+ '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。',
478
+ displayName: 'Ling-mini-2.0',
479
+ id: 'inclusionai/ling-mini-2.0',
480
+ maxOutput: 32_000,
481
+ pricing: {
482
+ units: [
483
+ { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
484
+ { name: 'textOutput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
485
+ ],
486
+ },
487
+ type: 'chat',
488
+ },
489
+ {
490
+ abilities: {
491
+ reasoning: true,
492
+ },
493
+ contextWindowTokens: 128_000,
494
+ 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。',
495
+ displayName: 'Ring-mini-2.0',
496
+ id: 'inclusionai/ring-mini-2.0',
497
+ maxOutput: 32_000,
498
+ pricing: {
499
+ units: [
500
+ { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
501
+ { name: 'textOutput', rate: 0.7, strategy: 'fixed', unit: 'millionTokens' },
502
+ ],
503
+ },
504
+ type: 'chat',
505
+ },
506
+ {
507
+ abilities: {
508
+ reasoning: true,
509
+ },
510
+ contextWindowTokens: 204_800,
511
+ 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。',
512
+ displayName: 'MiniMax M2',
513
+ id: 'minimax/minimax-m2',
514
+ maxOutput: 128_000,
515
+ pricing: {
516
+ units: [
517
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
518
+ { name: 'textOutput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
519
+ ],
520
+ },
521
+ type: 'chat',
522
+ },
523
+ {
524
+ contextWindowTokens: 256_000,
525
+ 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。',
526
+ displayName: 'KAT-Coder-Pro-V1 (Limited-time Free)',
527
+ id: 'kuaishou/kat-coder-pro-v1',
528
+ maxOutput: 32_000,
529
+ pricing: {
530
+ units: [
531
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
532
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
533
+ ],
534
+ },
535
+ type: 'chat',
536
+ },
537
+ {
538
+ abilities: {
539
+ functionCall: true,
540
+ reasoning: true,
541
+ search: true,
542
+ vision: true,
543
+ },
544
+ contextWindowTokens: 200_000,
545
+ description:
546
+ '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。',
547
+ displayName: 'Claude Haiku 4.5',
548
+ id: 'anthropic/claude-haiku-4.5',
549
+ maxOutput: 64_000,
550
+ pricing: {
551
+ units: [
552
+ { name: 'textInput', rate: 1, strategy: 'fixed', unit: 'millionTokens' },
553
+ { name: 'textOutput', rate: 5, strategy: 'fixed', unit: 'millionTokens' },
554
+ ],
555
+ },
556
+ settings: {
557
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
558
+ searchImpl: 'params',
559
+ },
560
+ type: 'chat',
561
+ },
562
+ {
563
+ abilities: {
564
+ reasoning: true,
565
+ },
566
+ contextWindowTokens: 128_000,
567
+ 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。',
568
+ displayName: 'DeepSeek-V3.2-Exp (Non-thinking Mode)',
569
+ id: 'deepseek/deepseek-chat',
570
+ maxOutput: 8000,
571
+ pricing: {
572
+ units: [
573
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
574
+ { name: 'textOutput', rate: 1.68, strategy: 'fixed', unit: 'millionTokens' },
575
+ ],
576
+ },
577
+ type: 'chat',
578
+ },
579
+ {
580
+ abilities: {
581
+ functionCall: true,
582
+ reasoning: true,
583
+ },
584
+ contextWindowTokens: 128_000,
585
+ description:
586
+ 'DeepSeek-V3 Thinking（reasoner）是 DeepSeek 的实验 reasoning 模型，适合高复杂度推理任务。',
587
+ displayName: 'DeepSeek-V3.2-Exp (Thinking Mode)',
588
+ id: 'deepseek/deepseek-reasoner',
589
+ maxOutput: 64_000,
590
+ pricing: {
591
+ units: [
592
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
593
+ { name: 'textOutput', rate: 0.42, strategy: 'fixed', unit: 'millionTokens' },
594
+ ],
595
+ },
596
+ type: 'chat',
597
+ },
598
+ {
599
+ abilities: {
600
+ reasoning: true,
601
+ },
602
+ contextWindowTokens: 64_000,
603
+ description: 'DeepSeek R1 0528 是 DeepSeek 的更新变体，注重开源可用与推理深度。',
604
+ displayName: 'DeepSeek R1 0528',
605
+ id: 'deepseek/deepseek-r1-0528',
606
+ maxOutput: 64_000,
607
+ pricing: {
608
+ units: [
609
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
610
+ { name: 'textOutput', rate: 2.23, strategy: 'fixed', unit: 'millionTokens' },
611
+ ],
612
+ },
613
+ type: 'chat',
614
+ },
615
+ {
616
+ abilities: {
617
+ functionCall: true,
618
+ reasoning: true,
619
+ search: true,
620
+ vision: true,
621
+ },
622
+ contextWindowTokens: 1_050_000,
623
+ 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。',
624
+ displayName: 'Gemini 2.5 Pro',
625
+ id: 'google/gemini-2.5-pro',
626
+ maxOutput: 65_530,
627
+ pricing: {
628
+ units: [
629
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
630
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
631
+ ],
632
+ },
633
+ settings: {
634
+ extendParams: ['thinkingBudget', 'urlContext'],
635
+ searchImpl: 'params',
636
+ searchProvider: 'google',
637
+ },
638
+ type: 'chat',
639
+ },
640
+ {
641
+ abilities: {
642
+ functionCall: true,
643
+ reasoning: true,
644
+ search: true,
645
+ vision: true,
646
+ },
647
+ contextWindowTokens: 1_050_000,
648
+ 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。',
649
+ displayName: 'Gemini 2.5 Pro Free',
650
+ id: 'google/gemini-2.5-pro-free',
651
+ maxOutput: 65_530,
652
+ pricing: {
653
+ units: [
654
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
655
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
656
+ ],
657
+ },
658
+ settings: {
659
+ extendParams: ['thinkingBudget', 'urlContext'],
660
+ searchImpl: 'params',
661
+ searchProvider: 'google',
662
+ },
663
+ type: 'chat',
664
+ },
665
+ {
666
+ abilities: {
667
+ functionCall: true,
668
+ reasoning: true,
669
+ search: true,
670
+ vision: true,
671
+ },
672
+ contextWindowTokens: 1_050_000,
673
+ 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。',
674
+ displayName: 'Gemini 2.5 Flash',
675
+ id: 'google/gemini-2.5-flash',
676
+ maxOutput: 65_530,
677
+ pricing: {
678
+ units: [
679
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
680
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
681
+ ],
682
+ },
683
+ settings: {
684
+ extendParams: ['thinkingBudget', 'urlContext'],
685
+ searchImpl: 'params',
686
+ searchProvider: 'google',
687
+ },
688
+ type: 'chat',
689
+ },
690
+ {
691
+ abilities: {
692
+ functionCall: true,
693
+ reasoning: true,
694
+ vision: true,
695
+ },
696
+ contextWindowTokens: 400_000,
697
+ description:
698
+ '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。',
699
+ displayName: 'GPT-5 Pro',
700
+ id: 'openai/gpt-5-pro',
701
+ maxOutput: 128_000,
702
+ pricing: {
703
+ units: [
704
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
705
+ { name: 'textOutput', rate: 120, strategy: 'fixed', unit: 'millionTokens' },
706
+ ],
707
+ },
708
+ settings: {
709
+ extendParams: ['textVerbosity'],
710
+ },
711
+ type: 'chat',
712
+ },
713
+ {
714
+ abilities: {
715
+ functionCall: true,
716
+ reasoning: true,
717
+ vision: true,
718
+ },
719
+ contextWindowTokens: 400_000,
720
+ description: 'GPT-5 是 OpenAI 的高性能模型，适用广泛的生产与研究任务。',
721
+ displayName: 'GPT-5',
722
+ id: 'openai/gpt-5',
723
+ maxOutput: 128_000,
724
+ pricing: {
725
+ units: [
726
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
727
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
728
+ ],
729
+ },
730
+ settings: {
731
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
732
+ },
733
+ type: 'chat',
734
+ },
735
+ {
736
+ abilities: {
737
+ functionCall: true,
738
+ reasoning: true,
739
+ },
740
+ contextWindowTokens: 128_000,
741
+ 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。',
742
+ displayName: 'GPT-5 Chat',
743
+ id: 'openai/gpt-5-chat',
744
+ maxOutput: 16_380,
745
+ pricing: {
746
+ units: [
747
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
748
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
749
+ ],
750
+ },
751
+ type: 'chat',
752
+ },
753
+ {
754
+ abilities: {
755
+ functionCall: true,
756
+ reasoning: true,
757
+ vision: true,
758
+ },
759
+ contextWindowTokens: 400_000,
760
+ description: 'GPT-5 Mini 是 GPT-5 家族的精简版，适用于低延迟低成本场景。',
761
+ displayName: 'GPT-5 Mini',
762
+ id: 'openai/gpt-5-mini',
763
+ maxOutput: 128_000,
764
+ pricing: {
765
+ units: [
766
+ { name: 'textInput', rate: 0.25, strategy: 'fixed', unit: 'millionTokens' },
767
+ { name: 'textOutput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
768
+ ],
769
+ },
770
+ settings: {
771
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
772
+ },
773
+ type: 'chat',
774
+ },
775
+ {
776
+ abilities: {
777
+ functionCall: true,
778
+ reasoning: true,
779
+ },
780
+ contextWindowTokens: 400_000,
781
+ 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。',
782
+ displayName: 'GPT-5 Nano',
783
+ id: 'openai/gpt-5-nano',
784
+ maxOutput: 128_000,
785
+ pricing: {
786
+ units: [
787
+ { name: 'textInput', rate: 0.05, strategy: 'fixed', unit: 'millionTokens' },
788
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
789
+ ],
790
+ },
791
+ settings: {
792
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
793
+ },
794
+ type: 'chat',
795
+ },
796
+ {
797
+ abilities: {
798
+ functionCall: true,
799
+ reasoning: true,
800
+ },
801
+ contextWindowTokens: 400_000,
802
+ 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。',
803
+ displayName: 'GPT-5 Codex',
804
+ id: 'openai/gpt-5-codex',
805
+ maxOutput: 128_000,
806
+ pricing: {
807
+ units: [
808
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
809
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
810
+ ],
811
+ },
812
+ settings: {
813
+ extendParams: ['gpt5ReasoningEffort', 'textVerbosity'],
814
+ },
815
+ type: 'chat',
816
+ },
817
+ {
818
+ abilities: {
819
+ functionCall: true,
820
+ reasoning: true,
821
+ },
822
+ contextWindowTokens: 1_050_000,
823
+ 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。',
824
+ displayName: 'GPT-4.1',
825
+ id: 'openai/gpt-4.1',
826
+ maxOutput: 32_770,
827
+ pricing: {
828
+ units: [
829
+ { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
830
+ { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
831
+ ],
832
+ },
833
+ type: 'chat',
834
+ },
835
+ {
836
+ abilities: {
837
+ reasoning: true,
838
+ },
839
+ contextWindowTokens: 1_050_000,
840
+ 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。',
841
+ displayName: 'GPT-4.1 Mini',
842
+ id: 'openai/gpt-4.1-mini',
843
+ maxOutput: 32_770,
844
+ pricing: {
845
+ units: [
846
+ { name: 'textInput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
847
+ { name: 'textOutput', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
848
+ ],
849
+ },
850
+ type: 'chat',
851
+ },
852
+ {
853
+ abilities: {
854
+ reasoning: true,
855
+ },
856
+ contextWindowTokens: 1_050_000,
857
+ 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。',
858
+ displayName: 'GPT-4.1 Nano',
859
+ id: 'openai/gpt-4.1-nano',
860
+ maxOutput: 32_770,
861
+ pricing: {
862
+ units: [
863
+ { name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
864
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
865
+ ],
866
+ },
867
+ type: 'chat',
868
+ },
869
+ {
870
+ abilities: {
871
+ reasoning: true,
872
+ },
873
+ contextWindowTokens: 262_144,
874
+ 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）。',
875
+ displayName: 'Qwen3 Max Thinking Preview',
876
+ id: 'qwen/qwen3-max-preview',
877
+ maxOutput: 65_540,
878
+ pricing: {
879
+ units: [
880
+ { name: 'textInput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
881
+ { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
882
+ ],
883
+ },
884
+ type: 'chat',
885
+ },
886
+ {
887
+ abilities: {
888
+ reasoning: true,
889
+ },
890
+ contextWindowTokens: 262_144,
891
+ 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。',
892
+ displayName: 'Qwen3 Max',
893
+ id: 'qwen/qwen3-max',
894
+ maxOutput: 65_540,
895
+ pricing: {
896
+ units: [
897
+ { name: 'textInput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
898
+ { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
899
+ ],
900
+ },
901
+ type: 'chat',
902
+ },
903
+ {
904
+ abilities: {
905
+ reasoning: true,
906
+ vision: true,
907
+ },
908
+ contextWindowTokens: 262_144,
909
+ 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。',
910
+ displayName: 'Qwen3-VL-Plus',
911
+ id: 'qwen/qwen3-vl-plus',
912
+ maxOutput: 32_770,
913
+ pricing: {
914
+ units: [
915
+ { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
916
+ { name: 'textOutput', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
917
+ ],
918
+ },
919
+ type: 'chat',
920
+ },
921
+ {
922
+ abilities: {
923
+ reasoning: true,
924
+ },
925
+ contextWindowTokens: 1_000_000,
926
+ description:
927
+ '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。',
928
+ displayName: 'Qwen3-Coder-Plus',
929
+ id: 'qwen/qwen3-coder-plus',
930
+ maxOutput: 65_540,
931
+ pricing: {
932
+ units: [
933
+ { name: 'textInput', rate: 1, strategy: 'fixed', unit: 'millionTokens' },
934
+ { name: 'textOutput', rate: 5, strategy: 'fixed', unit: 'millionTokens' },
935
+ ],
936
+ },
937
+ type: 'chat',
938
+ },
939
+ {
940
+ abilities: {
941
+ functionCall: true,
942
+ reasoning: true,
943
+ },
944
+ contextWindowTokens: 256_000,
945
+ 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。',
946
+ displayName: 'Qwen3-Coder',
947
+ id: 'qwen/qwen3-coder',
948
+ maxOutput: 261_000,
949
+ pricing: {
950
+ units: [
951
+ { name: 'textInput', rate: 1.25, strategy: 'fixed', unit: 'millionTokens' },
952
+ { name: 'textOutput', rate: 5.01, strategy: 'fixed', unit: 'millionTokens' },
953
+ ],
954
+ },
955
+ type: 'chat',
956
+ },
957
+ {
958
+ abilities: {
959
+ reasoning: true,
960
+ },
961
+ contextWindowTokens: 32_000,
962
+ description: 'Qwen3-14B 是 Qwen 系列的 14B 版本，适合常规推理与对话场景。',
963
+ displayName: 'Qwen3 14B',
964
+ id: 'qwen/qwen3-14b',
965
+ maxOutput: 32_000,
966
+ pricing: {
967
+ units: [
968
+ { name: 'textInput', rate: 0.14, strategy: 'fixed', unit: 'millionTokens' },
969
+ { name: 'textOutput', rate: 1.4, strategy: 'fixed', unit: 'millionTokens' },
970
+ ],
971
+ },
972
+ type: 'chat',
973
+ },
974
+ {
975
+ abilities: {
976
+ reasoning: true,
977
+ },
978
+ contextWindowTokens: 200_000,
979
+ 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。',
980
+ displayName: 'GLM 4.6',
981
+ id: 'z-ai/glm-4.6',
982
+ maxOutput: 128_000,
983
+ pricing: {
984
+ units: [
985
+ { name: 'textInput', rate: 0.35, strategy: 'fixed', unit: 'millionTokens' },
986
+ { name: 'textOutput', rate: 1.54, strategy: 'fixed', unit: 'millionTokens' },
987
+ ],
988
+ },
989
+ type: 'chat',
990
+ },
991
+ {
992
+ abilities: {
993
+ functionCall: true,
994
+ reasoning: true,
995
+ search: true,
996
+ vision: true,
997
+ },
998
+ contextWindowTokens: 200_000,
999
+ 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。',
1000
+ displayName: 'Claude Opus 4.1',
1001
+ id: 'anthropic/claude-opus-4.1',
1002
+ maxOutput: 32_000,
1003
+ pricing: {
1004
+ units: [
1005
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1006
+ { name: 'textOutput', rate: 75, strategy: 'fixed', unit: 'millionTokens' },
1007
+ ],
1008
+ },
1009
+ settings: {
1010
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1011
+ searchImpl: 'params',
1012
+ },
1013
+ type: 'chat',
1014
+ },
1015
+ {
1016
+ abilities: {
1017
+ functionCall: true,
1018
+ reasoning: true,
1019
+ search: true,
1020
+ vision: true,
1021
+ },
1022
+ contextWindowTokens: 200_000,
1023
+ 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。',
1024
+ displayName: 'Claude Opus 4',
1025
+ id: 'anthropic/claude-opus-4',
1026
+ maxOutput: 32_000,
1027
+ pricing: {
1028
+ units: [
1029
+ { name: 'textInput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1030
+ { name: 'textOutput', rate: 75, strategy: 'fixed', unit: 'millionTokens' },
1031
+ ],
1032
+ },
1033
+ settings: {
1034
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1035
+ searchImpl: 'params',
1036
+ },
1037
+ type: 'chat',
1038
+ },
1039
+ {
1040
+ abilities: {
1041
+ functionCall: true,
1042
+ imageOutput: true,
1043
+ reasoning: true,
1044
+ vision: true,
1045
+ },
1046
+ contextWindowTokens: 32_768,
1047
+ description:
1048
+ '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。',
1049
+ displayName: 'Gemini 2.5 Flash Image (Nano Banana)',
1050
+ id: 'google/gemini-2.5-flash-image',
1051
+ maxOutput: 8192,
1052
+ pricing: {
1053
+ approximatePricePerImage: 0.039,
1054
+ units: [
1055
+ { name: 'imageOutput', rate: 30, strategy: 'fixed', unit: 'millionTokens' },
1056
+ { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
1057
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1058
+ ],
1059
+ },
1060
+ type: 'chat',
1061
+ },
1062
+ {
1063
+ abilities: {
1064
+ functionCall: true,
1065
+ imageOutput: true,
1066
+ reasoning: true,
1067
+ vision: true,
1068
+ },
1069
+ contextWindowTokens: 32_768,
1070
+ description: 'Gemini 2.5 Flash Image 免费版，支持受限额度的多模态生成。',
1071
+ displayName: 'Gemini 2.5 Flash Image (Nano Banana) Free',
1072
+ id: 'google/gemini-2.5-flash-image-free',
1073
+ maxOutput: 8192,
1074
+ pricing: {
1075
+ units: [
1076
+ { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
1077
+ { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
1078
+ ],
1079
+ },
1080
+ type: 'chat',
1081
+ },
1082
+ {
1083
+ abilities: {
1084
+ functionCall: true,
1085
+ reasoning: true,
1086
+ search: true,
1087
+ vision: true,
1088
+ },
1089
+ contextWindowTokens: 200_000,
1090
+ 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。',
1091
+ displayName: 'Claude Sonnet 4.5',
1092
+ id: 'anthropic/claude-sonnet-4.5',
1093
+ maxOutput: 64_000,
1094
+ pricing: {
1095
+ units: [
1096
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1097
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1098
+ ],
1099
+ },
1100
+ settings: {
1101
+ extendParams: ['enableReasoning', 'reasoningBudgetToken'],
1102
+ searchImpl: 'params',
1103
+ },
1104
+ type: 'chat',
1105
+ },
1106
+ {
1107
+ abilities: {
1108
+ functionCall: true,
1109
+ reasoning: true,
1110
+ },
1111
+ contextWindowTokens: 1_000_000,
1112
+ description: 'Claude Sonnet 4 是 Anthropic 的混合推理版本，提供思维/非思维混合能力。',
1113
+ displayName: 'Claude Sonnet 4',
1114
+ id: 'anthropic/claude-sonnet-4',
1115
+ maxOutput: 64_000,
1116
+ pricing: {
1117
+ units: [
1118
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1119
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1120
+ ],
1121
+ },
1122
+ type: 'chat',
1123
+ },
1124
+ {
1125
+ abilities: {
1126
+ functionCall: true,
1127
+ reasoning: true,
1128
+ },
1129
+ contextWindowTokens: 200_000,
1130
+ description: 'OpenAI o4-mini 是 OpenAI 的小型高效推理模型，适合低延迟场景。',
1131
+ displayName: 'o4 Mini',
1132
+ id: 'openai/o4-mini',
1133
+ maxOutput: 100_000,
1134
+ pricing: {
1135
+ units: [
1136
+ { name: 'textInput', rate: 1.1, strategy: 'fixed', unit: 'millionTokens' },
1137
+ { name: 'textOutput', rate: 4.4, strategy: 'fixed', unit: 'millionTokens' },
1138
+ ],
1139
+ },
1140
+ settings: {
1141
+ extendParams: ['reasoningEffort'],
1142
+ },
1143
+ type: 'chat',
1144
+ },
1145
+ {
1146
+ abilities: {
1147
+ functionCall: true,
1148
+ reasoning: true,
1149
+ },
1150
+ contextWindowTokens: 128_000,
1151
+ description: 'GPT-4o 系列是 OpenAI 的 Omni 模型，支持文本 + 图片输入与文本输出。',
1152
+ displayName: 'GPT-4o',
1153
+ id: 'openai/gpt-4o',
1154
+ maxOutput: 16_380,
1155
+ pricing: {
1156
+ units: [
1157
+ { name: 'textInput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1158
+ { name: 'textOutput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
1159
+ ],
1160
+ },
1161
+ type: 'chat',
1162
+ },
1163
+ {
1164
+ abilities: {
1165
+ functionCall: true,
1166
+ reasoning: true,
1167
+ },
1168
+ contextWindowTokens: 128_000,
1169
+ 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。',
1170
+ displayName: 'GPT-4o-mini',
1171
+ id: 'openai/gpt-4o-mini',
1172
+ maxOutput: 16_380,
1173
+ pricing: {
1174
+ units: [
1175
+ { name: 'textInput', rate: 0.15, strategy: 'fixed', unit: 'millionTokens' },
1176
+ { name: 'textOutput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1177
+ ],
1178
+ },
1179
+ type: 'chat',
1180
+ },
1181
+ {
1182
+ abilities: {
1183
+ reasoning: true,
1184
+ },
1185
+ contextWindowTokens: 128_000,
1186
+ 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。',
1187
+ displayName: 'Kimi K2 0711',
1188
+ id: 'moonshotai/kimi-k2-0711',
1189
+ maxOutput: 32_000,
1190
+ pricing: {
1191
+ units: [
1192
+ { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
1193
+ { name: 'textOutput', rate: 2.23, strategy: 'fixed', unit: 'millionTokens' },
1194
+ ],
1195
+ },
1196
+ type: 'chat',
1197
+ },
1198
+ {
1199
+ abilities: {
1200
+ reasoning: true,
1201
+ },
1202
+ contextWindowTokens: 128_000,
1203
+ 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。',
1204
+ displayName: 'GLM 4.5 Air',
1205
+ id: 'z-ai/glm-4.5-air',
1206
+ maxOutput: 96_000,
1207
+ pricing: {
1208
+ units: [
1209
+ { name: 'textInput', rate: 0.11, strategy: 'fixed', unit: 'millionTokens' },
1210
+ { name: 'textOutput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
1211
+ ],
1212
+ },
1213
+ type: 'chat',
1214
+ },
1215
+ {
1216
+ abilities: {
1217
+ functionCall: true,
1218
+ reasoning: true,
1219
+ },
1220
+ contextWindowTokens: 200_000,
1221
+ description:
1222
+ '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。',
1223
+ displayName: 'Claude 3.5 Haiku',
1224
+ id: 'anthropic/claude-3.5-haiku',
1225
+ maxOutput: 8192,
1226
+ pricing: {
1227
+ units: [
1228
+ { name: 'textInput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
1229
+ { name: 'textOutput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
1230
+ ],
1231
+ },
1232
+ type: 'chat',
1233
+ },
1234
+ {
1235
+ abilities: {
1236
+ functionCall: true,
1237
+ reasoning: true,
1238
+ },
1239
+ contextWindowTokens: 200_000,
1240
+ description:
1241
+ '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。',
1242
+ displayName: 'Claude 3.5 Sonnet',
1243
+ id: 'anthropic/claude-3.5-sonnet',
1244
+ maxOutput: 8192,
1245
+ pricing: {
1246
+ units: [
1247
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1248
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1249
+ ],
1250
+ },
1251
+ type: 'chat',
1252
+ },
1253
+ {
1254
+ abilities: {
1255
+ functionCall: true,
1256
+ reasoning: true,
1257
+ },
1258
+ contextWindowTokens: 200_000,
1259
+ description:
1260
+ '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。',
1261
+ displayName: 'Claude 3.7 Sonnet',
1262
+ id: 'anthropic/claude-3.7-sonnet',
1263
+ maxOutput: 64_000,
1264
+ pricing: {
1265
+ units: [
1266
+ { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
1267
+ { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
1268
+ ],
1269
+ },
1270
+ type: 'chat',
1271
+ },
1272
+ {
1273
+ abilities: {
1274
+ functionCall: true,
1275
+ reasoning: true,
1276
+ search: true,
1277
+ },
1278
+ contextWindowTokens: 128_000,
1279
+ description:
1280
+ '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。',
1281
+ displayName: 'DeepSeek V3.1',
1282
+ id: 'deepseek/deepseek-chat-v3.1',
1283
+ maxOutput: 65_540,
1284
+ pricing: {
1285
+ units: [
1286
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1287
+ { name: 'textOutput', rate: 1.11, strategy: 'fixed', unit: 'millionTokens' },
1288
+ ],
1289
+ },
1290
+ type: 'chat',
1291
+ },
1292
+ {
1293
+ abilities: {
1294
+ functionCall: true,
1295
+ reasoning: true,
1296
+ vision: true,
1297
+ },
1298
+ contextWindowTokens: 1_050_000,
1299
+ 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。',
1300
+ displayName: 'Gemini 2.0 Flash',
1301
+ id: 'google/gemini-2.0-flash',
1302
+ maxOutput: 8192,
1303
+ pricing: {
1304
+ units: [
1305
+ { name: 'textInput', rate: 0.15, strategy: 'fixed', unit: 'millionTokens' },
1306
+ { name: 'textOutput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1307
+ ],
1308
+ },
1309
+ type: 'chat',
1310
+ },
1311
+ {
1312
+ abilities: {
1313
+ functionCall: true,
1314
+ reasoning: true,
1315
+ vision: true,
1316
+ },
1317
+ contextWindowTokens: 1_050_000,
1318
+ description:
1319
+ '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。',
1320
+ displayName: 'Gemini 2.0 Flash Lite',
1321
+ id: 'google/gemini-2.0-flash-lite-001',
1322
+ maxOutput: 8192,
1323
+ pricing: {
1324
+ units: [
1325
+ { name: 'textInput', rate: 0.075, strategy: 'fixed', unit: 'millionTokens' },
1326
+ { name: 'textOutput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
1327
+ ],
1328
+ },
1329
+ type: 'chat',
1330
+ },
1331
+ {
1332
+ abilities: {
1333
+ functionCall: true,
1334
+ reasoning: true,
1335
+ vision: true,
1336
+ },
1337
+ contextWindowTokens: 1_050_000,
1338
+ description:
1339
+ '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。',
1340
+ displayName: 'Gemini 2.5 Flash Lite',
1341
+ id: 'google/gemini-2.5-flash-lite',
1342
+ maxOutput: 65_530,
1343
+ pricing: {
1344
+ units: [
1345
+ { name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
1346
+ { name: 'textOutput', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
1347
+ ],
1348
+ },
1349
+ type: 'chat',
1350
+ },
1351
+ {
1352
+ abilities: {
1353
+ reasoning: true,
1354
+ },
1355
+ contextWindowTokens: 262_144,
1356
+ 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。',
1357
+ displayName: 'Kimi K2 0905',
1358
+ id: 'moonshotai/kimi-k2-0905',
1359
+ maxOutput: 262_144,
1360
+ pricing: {
1361
+ units: [
1362
+ { name: 'textInput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
1363
+ { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
1364
+ ],
1365
+ },
1366
+ type: 'chat',
1367
+ },
1368
+ {
1369
+ abilities: {
1370
+ reasoning: true,
1371
+ },
1372
+ contextWindowTokens: 256_000,
1373
+ description:
1374
+ '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。',
1375
+ displayName: 'Qwen3 235B A22B Instruct 2507',
1376
+ id: 'qwen/qwen3-235b-a22b-2507',
1377
+ maxOutput: 262_100,
1378
+ pricing: {
1379
+ units: [
1380
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1381
+ { name: 'textOutput', rate: 1.11, strategy: 'fixed', unit: 'millionTokens' },
1382
+ ],
1383
+ },
1384
+ type: 'chat',
1385
+ },
1386
+ {
1387
+ abilities: {
1388
+ reasoning: true,
1389
+ },
1390
+ contextWindowTokens: 256_000,
1391
+ description:
1392
+ '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。',
1393
+ displayName: 'Qwen3 235B A22B Thinking 2507',
1394
+ id: 'qwen/qwen3-235b-a22b-thinking-2507',
1395
+ maxOutput: 262_100,
1396
+ pricing: {
1397
+ units: [
1398
+ { name: 'textInput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
1399
+ { name: 'textOutput', rate: 2.78, strategy: 'fixed', unit: 'millionTokens' },
1400
+ ],
1401
+ },
1402
+ type: 'chat',
1403
+ },
1404
+ {
1405
+ abilities: { reasoning: true },
1406
+ contextWindowTokens: 128_000,
1407
+ 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。',
1408
+ displayName: 'GLM 4.5',
1409
+ id: 'z-ai/glm-4.5',
1410
+ maxOutput: 96_000,
1411
+ pricing: {
1412
+ units: [
1413
+ { name: 'textInput', rate: 0.35, strategy: 'fixed', unit: 'millionTokens' },
1414
+ { name: 'textOutput', rate: 1.54, strategy: 'fixed', unit: 'millionTokens' },
1415
+ ],
1416
+ },
1417
+ type: 'chat',
1418
+ },
1419
+ ];
1420
+
1421
+ export const allModels = [...zenmuxChatModels];
1422
+
1423
+ export default allModels;