@lobehub/chat 1.80.0 → 1.80.2

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. package/CHANGELOG.md +58 -0
  2. package/changelog/v1.json +21 -0
  3. package/docs/development/basic/feature-development.mdx +370 -619
  4. package/docs/development/basic/feature-development.zh-CN.mdx +368 -611
  5. package/package.json +1 -1
  6. package/src/app/[variants]/oauth/consent/[uid]/Client.tsx +36 -23
  7. package/src/app/[variants]/oauth/consent/[uid]/page.tsx +2 -0
  8. package/src/config/aiModels/azure.ts +79 -1
  9. package/src/config/aiModels/azureai.ts +181 -0
  10. package/src/config/aiModels/google.ts +36 -2
  11. package/src/config/aiModels/groq.ts +31 -3
  12. package/src/config/aiModels/hunyuan.ts +54 -18
  13. package/src/config/aiModels/moonshot.ts +17 -17
  14. package/src/config/aiModels/novita.ts +25 -30
  15. package/src/config/aiModels/siliconcloud.ts +80 -2
  16. package/src/config/aiModels/stepfun.ts +40 -31
  17. package/src/config/aiModels/tencentcloud.ts +7 -6
  18. package/src/config/aiModels/volcengine.ts +1 -0
  19. package/src/config/aiModels/zhipu.ts +91 -27
  20. package/src/const/settings/knowledge.ts +2 -2
  21. package/src/features/ChatInput/ActionBar/Upload/ClientMode.tsx +7 -6
  22. package/src/hooks/useModelSupportFiles.ts +15 -0
  23. package/src/libs/agent-runtime/stepfun/index.ts +7 -1
  24. package/src/libs/agent-runtime/zhipu/index.ts +17 -10
  25. package/src/libs/oidc-provider/config.ts +0 -3
  26. package/src/libs/trpc/edge/index.ts +0 -4
  27. package/src/libs/trpc/lambda/context.ts +90 -6
  28. package/src/libs/trpc/lambda/index.ts +2 -1
  29. package/src/libs/trpc/lambda/middleware/oidcAuth.ts +14 -0
  30. package/src/libs/trpc/middleware/userAuth.ts +2 -4
  31. package/src/server/services/oidc/index.ts +71 -0
  32. package/src/store/aiInfra/slices/aiModel/selectors.ts +7 -0
  33. package/src/utils/parseModels.test.ts +19 -3
  34. package/src/utils/server/__tests__/auth.test.ts +45 -1
  35. package/src/utils/server/auth.ts +26 -2
  36. package/docs/development/basic/feature-development-new.mdx +0 -465
  37. package/docs/development/basic/feature-development-new.zh-CN.mdx +0 -465
@@ -1,5 +1,5 @@
1
1
  import { AIChatModelCard } from '@/types/aiModel';
2
-
2
+ // https://platform.moonshot.cn/docs/pricing/chat
3
3
  const moonshotChatModels: AIChatModelCard[] = [
4
4
  {
5
5
  abilities: {
@@ -16,8 +16,8 @@ const moonshotChatModels: AIChatModelCard[] = [
16
16
  pricing: {
17
17
  cachedInput: 1,
18
18
  currency: 'CNY',
19
- input: 60,
20
- output: 60,
19
+ input: 10,
20
+ output: 30,
21
21
  },
22
22
  releasedAt: '2025-02-17',
23
23
  settings: {
@@ -37,8 +37,8 @@ const moonshotChatModels: AIChatModelCard[] = [
37
37
  id: 'moonshot-v1-auto',
38
38
  pricing: {
39
39
  currency: 'CNY',
40
- input: 60,
41
- output: 60,
40
+ input: 10,
41
+ output: 30,
42
42
  },
43
43
  settings: {
44
44
  searchImpl: 'params',
@@ -57,8 +57,8 @@ const moonshotChatModels: AIChatModelCard[] = [
57
57
  id: 'moonshot-v1-8k',
58
58
  pricing: {
59
59
  currency: 'CNY',
60
- input: 12,
61
- output: 12,
60
+ input: 2,
61
+ output: 10,
62
62
  },
63
63
  settings: {
64
64
  searchImpl: 'params',
@@ -77,8 +77,8 @@ const moonshotChatModels: AIChatModelCard[] = [
77
77
  id: 'moonshot-v1-32k',
78
78
  pricing: {
79
79
  currency: 'CNY',
80
- input: 24,
81
- output: 24,
80
+ input: 5,
81
+ output: 20,
82
82
  },
83
83
  settings: {
84
84
  searchImpl: 'params',
@@ -97,8 +97,8 @@ const moonshotChatModels: AIChatModelCard[] = [
97
97
  id: 'moonshot-v1-128k',
98
98
  pricing: {
99
99
  currency: 'CNY',
100
- input: 60,
101
- output: 60,
100
+ input: 10,
101
+ output: 30,
102
102
  },
103
103
  settings: {
104
104
  searchImpl: 'params',
@@ -118,8 +118,8 @@ const moonshotChatModels: AIChatModelCard[] = [
118
118
  id: 'moonshot-v1-8k-vision-preview',
119
119
  pricing: {
120
120
  currency: 'CNY',
121
- input: 12,
122
- output: 12,
121
+ input: 2,
122
+ output: 10,
123
123
  },
124
124
  releasedAt: '2025-01-14',
125
125
  settings: {
@@ -140,8 +140,8 @@ const moonshotChatModels: AIChatModelCard[] = [
140
140
  id: 'moonshot-v1-32k-vision-preview',
141
141
  pricing: {
142
142
  currency: 'CNY',
143
- input: 24,
144
- output: 24,
143
+ input: 5,
144
+ output: 20,
145
145
  },
146
146
  releasedAt: '2025-01-14',
147
147
  settings: {
@@ -162,8 +162,8 @@ const moonshotChatModels: AIChatModelCard[] = [
162
162
  id: 'moonshot-v1-128k-vision-preview',
163
163
  pricing: {
164
164
  currency: 'CNY',
165
- input: 60,
166
- output: 60,
165
+ input: 10,
166
+ output: 30,
167
167
  },
168
168
  releasedAt: '2025-01-14',
169
169
  settings: {
@@ -1,26 +1,36 @@
1
1
  import { AIChatModelCard } from '@/types/aiModel';
2
2
 
3
3
  // https://novita.ai/pricing
4
-
5
4
  const novitaChatModels: AIChatModelCard[] = [
6
5
  {
7
6
  contextWindowTokens: 131_072,
8
- displayName: 'Llama 3.3 70B Instruct',
7
+ displayName: 'Llama 4 Scout 17B Instruct',
9
8
  enabled: true,
10
- id: 'meta-llama/llama-3.3-70b-instruct',
9
+ id: 'meta-llama/llama-4-scout-17b-16e-instruct',
11
10
  pricing: {
12
- input: 0.13,
13
- output: 0.39,
11
+ input: 0.1,
12
+ output: 0.5,
13
+ },
14
+ type: 'chat',
15
+ },
16
+ {
17
+ contextWindowTokens: 1_048_576,
18
+ displayName: 'Llama 4 Maverick 17B Instruct',
19
+ enabled: true,
20
+ id: 'meta-llama/llama-4-maverick-17b-128e-instruct-fp8',
21
+ pricing: {
22
+ input: 0.2,
23
+ output: 0.85,
14
24
  },
15
25
  type: 'chat',
16
26
  },
17
27
  {
18
28
  contextWindowTokens: 16_384,
19
- description: 'Llama 3.1 8B Instruct 是 Meta 推出的最新版本,优化了高质量对话场景,表现优于许多领先的闭源模型。',
29
+ description: 'Llama 3.1 8B Instruct 优化了高质量对话场景,表现优于许多领先的闭源模型。',
20
30
  displayName: 'Llama 3.1 8B Instruct',
21
31
  id: 'meta-llama/llama-3.1-8b-instruct',
22
32
  pricing: {
23
- input: 0.05,
33
+ input: 0.02,
24
34
  output: 0.05,
25
35
  },
26
36
  type: 'chat',
@@ -73,7 +83,6 @@ const novitaChatModels: AIChatModelCard[] = [
73
83
  contextWindowTokens: 32_000,
74
84
  description: 'Gemma 3 27B 是谷歌的一款开源语言模型,以其在效率和性能方面设立了新的标准。',
75
85
  displayName: 'Gemma 3 27B',
76
- enabled: true,
77
86
  id: 'google/gemma-3-27b-it',
78
87
  pricing: {
79
88
  input: 0.2,
@@ -190,7 +199,7 @@ const novitaChatModels: AIChatModelCard[] = [
190
199
  enabled: true,
191
200
  id: 'deepseek/deepseek-v3-0324',
192
201
  pricing: {
193
- input: 0.4,
202
+ input: 0.37,
194
203
  output: 1.3,
195
204
  },
196
205
  type: 'chat',
@@ -318,18 +327,12 @@ const novitaChatModels: AIChatModelCard[] = [
318
327
  type: 'chat',
319
328
  },
320
329
  {
321
- contextWindowTokens: 32_768,
322
- displayName: 'Qwen 2 VL 72B Instruct',
323
- id: 'qwen/qwen-2-vl-72b-instruct',
324
- pricing: {
325
- input: 0.45,
326
- output: 0.45,
330
+ abilities: {
331
+ vision: true,
327
332
  },
328
- type: 'chat',
329
- },
330
- {
331
333
  contextWindowTokens: 96_000,
332
334
  displayName: 'Qwen 2.5 VL 72B Instruct',
335
+ enabled: true,
333
336
  id: 'qwen/qwen2.5-vl-72b-instruct',
334
337
  pricing: {
335
338
  input: 0.8,
@@ -378,7 +381,7 @@ const novitaChatModels: AIChatModelCard[] = [
378
381
  type: 'chat',
379
382
  },
380
383
  {
381
- contextWindowTokens: 16_000,
384
+ contextWindowTokens: 8192,
382
385
  displayName: 'L31 70B Euryale v2.2',
383
386
  id: 'sao10k/l31-70b-euryale-v2.2',
384
387
  pricing: {
@@ -387,22 +390,13 @@ const novitaChatModels: AIChatModelCard[] = [
387
390
  },
388
391
  type: 'chat',
389
392
  },
390
- {
391
- contextWindowTokens: 32_768,
392
- displayName: 'Qwen 2 7B Instruct',
393
- id: 'qwen/qwen-2-7b-instruct',
394
- pricing: {
395
- input: 0.054,
396
- output: 0.054,
397
- },
398
- type: 'chat',
399
- },
400
- {
393
+ {
401
394
  abilities: {
402
395
  reasoning: true,
403
396
  },
404
397
  contextWindowTokens: 32_768,
405
398
  displayName: 'QwQ 32B',
399
+ enabled: true,
406
400
  id: 'qwen/qwq-32b',
407
401
  pricing: {
408
402
  input: 0.18,
@@ -410,6 +404,7 @@ const novitaChatModels: AIChatModelCard[] = [
410
404
  },
411
405
  type: 'chat',
412
406
  },
407
+
413
408
  ];
414
409
 
415
410
  export const allModels = [...novitaChatModels];
@@ -3,6 +3,85 @@ import { AIChatModelCard } from '@/types/aiModel';
3
3
  // https://siliconflow.cn/zh-cn/models
4
4
 
5
5
  const siliconcloudChatModels: AIChatModelCard[] = [
6
+ {
7
+ abilities: {
8
+ reasoning: true,
9
+ },
10
+ contextWindowTokens: 32_768,
11
+ description:
12
+ 'GLM-Z1-Rumination-32B-0414 是一个具有沉思能力的深度推理模型(与 OpenAI 的 Deep Research 对标)。与典型的深度思考模型不同,沉思模型采用更长时间的深度思考来解决更开放和复杂的问题。',
13
+ displayName: 'GLM-Z1-Rumination 32B 0414',
14
+ enabled: true,
15
+ id: 'THUDM/GLM-Z1-Rumination-32B-0414',
16
+ pricing: {
17
+ currency: 'CNY',
18
+ input: 0.5,
19
+ output: 0.5,
20
+ },
21
+ type: 'chat',
22
+ },
23
+ {
24
+ abilities: {
25
+ reasoning: true,
26
+ },
27
+ contextWindowTokens: 32_768,
28
+ description:
29
+ 'GLM-Z1-32B-0414 是一个具有深度思考能力的推理模型。该模型基于 GLM-4-32B-0414 通过冷启动和扩展强化学习开发,并在数学、代码和逻辑任务上进行了进一步训练。与基础模型相比,GLM-Z1-32B-0414 显著提升了数学能力和解决复杂任务的能力。',
30
+ displayName: 'GLM-Z1 32B 0414',
31
+ enabled: true,
32
+ id: 'THUDM/GLM-Z1-32B-0414',
33
+ pricing: {
34
+ currency: 'CNY',
35
+ input: 0.5,
36
+ output: 0.5,
37
+ },
38
+ type: 'chat',
39
+ },
40
+ {
41
+ abilities: {
42
+ reasoning: true,
43
+ },
44
+ contextWindowTokens: 32_768,
45
+ description:
46
+ 'GLM-Z1-9B-0414 是 GLM 系列的小型模型,仅有 90 亿参数,但保持了开源传统的同时展现出惊人的能力。尽管规模较小,该模型在数学推理和通用任务上仍表现出色,其总体性能在同等规模的开源模型中已处于领先水平。',
47
+ displayName: 'GLM-Z1 9B 0414',
48
+ enabled: true,
49
+ id: 'THUDM/GLM-Z1-9B-0414',
50
+ pricing: {
51
+ currency: 'CNY',
52
+ input: 0,
53
+ output: 0,
54
+ },
55
+ type: 'chat',
56
+ },
57
+ {
58
+ contextWindowTokens: 32_768,
59
+ description:
60
+ 'GLM-4-32B-0414 是 GLM 系列的新一代开源模型,拥有 320 亿参数。该模型性能可与 OpenAI 的 GPT 系列和 DeepSeek 的 V3/R1 系列相媲美。',
61
+ displayName: 'GLM-4 32B 0414',
62
+ enabled: true,
63
+ id: 'THUDM/GLM-4-32B-0414',
64
+ pricing: {
65
+ currency: 'CNY',
66
+ input: 0.5,
67
+ output: 0.5,
68
+ },
69
+ type: 'chat',
70
+ },
71
+ {
72
+ contextWindowTokens: 32_768,
73
+ description:
74
+ 'GLM-4-9B-0414 是 GLM 系列的小型模型,拥有 90 亿参数。该模型继承了 GLM-4-32B 系列的技术特点,但提供了更轻量级的部署选择。尽管规模较小,GLM-4-9B-0414 仍在代码生成、网页设计、SVG 图形生成和基于搜索的写作等任务上展现出色能力。',
75
+ displayName: 'GLM-4 9B 0414',
76
+ enabled: true,
77
+ id: 'THUDM/GLM-4-9B-0414',
78
+ pricing: {
79
+ currency: 'CNY',
80
+ input: 0,
81
+ output: 0,
82
+ },
83
+ type: 'chat',
84
+ },
6
85
  {
7
86
  abilities: {
8
87
  functionCall: true,
@@ -29,6 +108,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
29
108
  description:
30
109
  'DeepSeek-V3 是一款拥有 6710 亿参数的混合专家(MoE)语言模型,采用多头潜在注意力(MLA)和 DeepSeekMoE 架构,结合无辅助损失的负载平衡策略,优化推理和训练效率。通过在 14.8 万亿高质量tokens上预训练,并进行监督微调和强化学习,DeepSeek-V3 在性能上超越其他开源模型,接近领先闭源模型。',
31
110
  displayName: 'DeepSeek V3',
111
+ enabled: true,
32
112
  id: 'deepseek-ai/DeepSeek-V3',
33
113
  pricing: {
34
114
  currency: 'CNY',
@@ -369,7 +449,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
369
449
  description:
370
450
  'Qwen2.5-72B-Instruct 是阿里云发布的最新大语言模型系列之一。该 72B 模型在编码和数学等领域具有显著改进的能力。它支持长达 128K tokens 的输入,可以生成超过 8K tokens 的长文本。该模型还提供了多语言支持,覆盖超过 29 种语言,包括中文、英文等。模型在指令跟随、理解结构化数据以及生成结构化输出(尤其是 JSON)方面都有显著提升',
371
451
  displayName: 'Qwen2.5 72B Instruct 128K',
372
- enabled: true,
373
452
  id: 'Qwen/Qwen2.5-72B-Instruct-128K',
374
453
  pricing: {
375
454
  currency: 'CNY',
@@ -541,7 +620,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
541
620
  description:
542
621
  'Qwen2.5-VL 是 Qwen2.5 系列中的视觉语言模型。该模型在多方面有显著提升:具备更强的视觉理解能力,能够识别常见物体、分析文本、图表和布局;作为视觉代理能够推理并动态指导工具使用;支持理解超过 1 小时的长视频并捕捉关键事件;能够通过生成边界框或点准确定位图像中的物体;支持生成结构化输出,尤其适用于发票、表格等扫描数据。',
543
622
  displayName: 'Qwen2.5 VL 72B Instruct',
544
- enabled: true,
545
623
  id: 'Qwen/Qwen2.5-VL-72B-Instruct',
546
624
  pricing: {
547
625
  currency: 'CNY',
@@ -3,6 +3,27 @@ import { AIChatModelCard } from '@/types/aiModel';
3
3
  // https://platform.stepfun.com/docs/pricing/details
4
4
 
5
5
  const stepfunChatModels: AIChatModelCard[] = [
6
+ {
7
+ abilities: {
8
+ // functionCall: true,
9
+ reasoning: true,
10
+ // search: true,
11
+ vision: true,
12
+ },
13
+ contextWindowTokens: 100_000,
14
+ description: '该模型是拥有强大的图像理解能力的推理大模型,能够处理图像和文字信息,经过深度思考后输出文本生成文本内容。该模型在视觉推理领域表现突出,同时拥有第一梯队的数学、代码、文本推理能力。上下文长度为100k。',
15
+ displayName: 'Step R1 V Mini',
16
+ id: 'step-r1-v-mini',
17
+ pricing: {
18
+ currency: 'CNY',
19
+ input: 2.5,
20
+ output: 8,
21
+ },
22
+ // settings: {
23
+ // searchImpl: 'params',
24
+ // },
25
+ type: 'chat',
26
+ },
6
27
  {
7
28
  abilities: {
8
29
  functionCall: true,
@@ -11,8 +32,7 @@ const stepfunChatModels: AIChatModelCard[] = [
11
32
  contextWindowTokens: 8000,
12
33
  description: '高速模型,适合实时对话。',
13
34
  displayName: 'Step 1 Flash',
14
- enabled: true,
15
- id: 'step-1-flash',
35
+ id: 'step-1-flash', // 将在2025年4月30日下线
16
36
  pricing: {
17
37
  currency: 'CNY',
18
38
  input: 1,
@@ -31,7 +51,6 @@ const stepfunChatModels: AIChatModelCard[] = [
31
51
  contextWindowTokens: 8000,
32
52
  description: '小型模型,适合轻量级任务。',
33
53
  displayName: 'Step 1 8K',
34
- enabled: true,
35
54
  id: 'step-1-8k',
36
55
  pricing: {
37
56
  currency: 'CNY',
@@ -51,7 +70,6 @@ const stepfunChatModels: AIChatModelCard[] = [
51
70
  contextWindowTokens: 32_000,
52
71
  description: '支持中等长度的对话,适用于多种应用场景。',
53
72
  displayName: 'Step 1 32K',
54
- enabled: true,
55
73
  id: 'step-1-32k',
56
74
  pricing: {
57
75
  currency: 'CNY',
@@ -71,8 +89,7 @@ const stepfunChatModels: AIChatModelCard[] = [
71
89
  contextWindowTokens: 128_000,
72
90
  description: '平衡性能与成本,适合一般场景。',
73
91
  displayName: 'Step 1 128K',
74
- enabled: true,
75
- id: 'step-1-128k',
92
+ id: 'step-1-128k', // 将在2025年4月30日下线
76
93
  pricing: {
77
94
  currency: 'CNY',
78
95
  input: 40,
@@ -107,16 +124,18 @@ const stepfunChatModels: AIChatModelCard[] = [
107
124
  functionCall: true,
108
125
  search: true,
109
126
  },
110
- contextWindowTokens: 16_000,
111
- description: '支持大规模上下文交互,适合复杂对话场景。',
112
- displayName: 'Step 2 16K',
127
+ contextWindowTokens: 8000,
128
+ description:
129
+ '基于新一代自研Attention架构MFA的极速大模型,用极低成本达到和step1类似的效果,同时保持了更高的吞吐和更快响应时延。能够处理通用任务,在代码能力上具备特长。',
130
+ displayName: 'Step 2 Mini',
113
131
  enabled: true,
114
- id: 'step-2-16k',
115
- pricing: {
132
+ id: 'step-2-mini',
133
+ pricing: {
116
134
  currency: 'CNY',
117
- input: 38,
118
- output: 120,
135
+ input: 1,
136
+ output: 2,
119
137
  },
138
+ releasedAt: '2025-01-14',
120
139
  settings: {
121
140
  searchImpl: 'params',
122
141
  },
@@ -127,18 +146,15 @@ const stepfunChatModels: AIChatModelCard[] = [
127
146
  functionCall: true,
128
147
  search: true,
129
148
  },
130
- contextWindowTokens: 8000,
131
- description:
132
- '基于新一代自研Attention架构MFA的极速大模型,用极低成本达到和step1类似的效果,同时保持了更高的吞吐和更快响应时延。能够处理通用任务,在代码能力上具备特长。',
133
- displayName: 'Step 2 Mini',
134
- enabled: true,
135
- id: 'step-2-mini',
136
- pricing: {
149
+ contextWindowTokens: 16_000,
150
+ description: '支持大规模上下文交互,适合复杂对话场景。',
151
+ displayName: 'Step 2 16K',
152
+ id: 'step-2-16k',
153
+ pricing: {
137
154
  currency: 'CNY',
138
- input: 1,
139
- output: 2,
155
+ input: 38,
156
+ output: 120,
140
157
  },
141
- releasedAt: '2025-01-14',
142
158
  settings: {
143
159
  searchImpl: 'params',
144
160
  },
@@ -153,7 +169,7 @@ const stepfunChatModels: AIChatModelCard[] = [
153
169
  description: 'step-2模型的实验版本,包含最新的特性,滚动更新中。不推荐在正式生产环境使用。',
154
170
  displayName: 'Step 2 16K Exp',
155
171
  enabled: true,
156
- id: 'step-2-16k',
172
+ id: 'step-2-16k-exp',
157
173
  pricing: {
158
174
  currency: 'CNY',
159
175
  input: 38,
@@ -174,7 +190,6 @@ const stepfunChatModels: AIChatModelCard[] = [
174
190
  contextWindowTokens: 8000,
175
191
  description: '小型视觉模型,适合基本的图文任务。',
176
192
  displayName: 'Step 1V 8K',
177
- enabled: true,
178
193
  id: 'step-1v-8k',
179
194
  pricing: {
180
195
  currency: 'CNY',
@@ -208,8 +223,6 @@ const stepfunChatModels: AIChatModelCard[] = [
208
223
  },
209
224
  {
210
225
  abilities: {
211
- functionCall: true,
212
- search: true,
213
226
  vision: true,
214
227
  },
215
228
  contextWindowTokens: 32_000,
@@ -223,9 +236,6 @@ const stepfunChatModels: AIChatModelCard[] = [
223
236
  output: 70,
224
237
  },
225
238
  releasedAt: '2025-01-22',
226
- settings: {
227
- searchImpl: 'params',
228
- },
229
239
  type: 'chat',
230
240
  },
231
241
  {
@@ -235,7 +245,6 @@ const stepfunChatModels: AIChatModelCard[] = [
235
245
  contextWindowTokens: 32_000,
236
246
  description: '该模型拥有强大的视频理解能力。',
237
247
  displayName: 'Step 1.5V Mini',
238
- enabled: true,
239
248
  id: 'step-1.5v-mini',
240
249
  pricing: {
241
250
  currency: 'CNY',
@@ -1,5 +1,5 @@
1
1
  import { AIChatModelCard } from '@/types/aiModel';
2
-
2
+ // https://cloud.tencent.com/document/product/1772/115969
3
3
  const tencentCloudChatModels: AIChatModelCard[] = [
4
4
  {
5
5
  abilities: {
@@ -7,11 +7,11 @@ const tencentCloudChatModels: AIChatModelCard[] = [
7
7
  },
8
8
  contextWindowTokens: 65_536,
9
9
  description:
10
- 'DeepSeek-R1 是一款强化学习(RL)驱动的推理模型,解决了模型中的重复性和可读性问题。在 RL 之前,DeepSeek-R1 引入了冷启动数据,进一步优化了推理性能。它在数学、代码和推理任务中与 OpenAI-o1 表现相当,并且通过精心设计的训练方法,提升了整体效果。',
10
+ 'DeepSeek-R1 为671B 模型,使用强化学习训练,推理过程包含大量反思和验证,思维链长度可达数万字。 该系列模型在数学、代码以及各种复杂逻辑推理任务上推理效果优异,并为用户展现了完整的思考过程。',
11
11
  displayName: 'DeepSeek R1',
12
12
  enabled: true,
13
13
  id: 'deepseek-r1',
14
- maxOutput: 8192,
14
+ maxOutput: 16_000,
15
15
  pricing: {
16
16
  currency: 'CNY',
17
17
  input: 4,
@@ -23,9 +23,10 @@ const tencentCloudChatModels: AIChatModelCard[] = [
23
23
  contextWindowTokens: 65_536,
24
24
  description:
25
25
  'DeepSeek-V3-0324 为671B 参数 MoE 模型,在编程与技术能力、上下文理解与长文本处理等方面优势突出。',
26
- displayName: 'DeepSeek-V3-0324',
26
+ displayName: 'DeepSeek V3 0324',
27
27
  enabled: true,
28
28
  id: 'deepseek-v3-0324',
29
+ maxOutput: 16_000,
29
30
  pricing: {
30
31
  currency: 'CNY',
31
32
  input: 2,
@@ -36,10 +37,10 @@ const tencentCloudChatModels: AIChatModelCard[] = [
36
37
  {
37
38
  contextWindowTokens: 65_536,
38
39
  description:
39
- 'DeepSeek-V3 是一款拥有 6710 亿参数的混合专家(MoE)语言模型,采用多头潜在注意力(MLA)和 DeepSeekMoE 架构,结合无辅助损失的负载平衡策略,优化推理和训练效率。通过在 14.8 万亿高质量tokens上预训练,并进行监督微调和强化学习,DeepSeek-V3 在性能上超越其他开源模型,接近领先闭源模型。',
40
+ 'DeepSeek-V3 为671B 参数 MoE 模型,在百科知识、数学推理等多项任务上优势突出。',
40
41
  displayName: 'DeepSeek V3',
41
- enabled: true,
42
42
  id: 'deepseek-v3',
43
+ maxOutput: 16_000,
43
44
  pricing: {
44
45
  currency: 'CNY',
45
46
  input: 2,
@@ -149,6 +149,7 @@ const doubaoChatModels: AIChatModelCard[] = [
149
149
  },
150
150
  {
151
151
  abilities: {
152
+ functionCall: true,
152
153
  vision: true,
153
154
  },
154
155
  config: {