@lobehub/chat 1.97.16 → 1.98.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +50 -0
- package/apps/desktop/package.json +8 -5
- package/apps/desktop/src/main/const/store.ts +12 -0
- package/apps/desktop/src/main/controllers/NetworkProxyCtr.ts +172 -0
- package/apps/desktop/src/main/controllers/__tests__/NetworkProxyCtr.test.ts +401 -0
- package/apps/desktop/src/main/core/Browser.ts +2 -0
- package/apps/desktop/src/main/modules/networkProxy/dispatcher.ts +116 -0
- package/apps/desktop/src/main/modules/networkProxy/index.ts +6 -0
- package/apps/desktop/src/main/modules/networkProxy/tester.ts +163 -0
- package/apps/desktop/src/main/modules/networkProxy/urlBuilder.ts +25 -0
- package/apps/desktop/src/main/modules/networkProxy/validator.ts +80 -0
- package/apps/desktop/src/main/types/store.ts +2 -1
- package/apps/desktop/src/main/utils/logger.ts +2 -1
- package/changelog/v1.json +18 -0
- package/locales/ar/electron.json +39 -0
- package/locales/ar/setting.json +1 -0
- package/locales/bg-BG/electron.json +39 -0
- package/locales/bg-BG/setting.json +1 -0
- package/locales/de-DE/electron.json +39 -0
- package/locales/de-DE/setting.json +1 -0
- package/locales/en-US/electron.json +39 -0
- package/locales/en-US/setting.json +1 -0
- package/locales/es-ES/electron.json +39 -0
- package/locales/es-ES/setting.json +1 -0
- package/locales/fa-IR/electron.json +39 -0
- package/locales/fa-IR/setting.json +1 -0
- package/locales/fr-FR/electron.json +39 -0
- package/locales/fr-FR/setting.json +1 -0
- package/locales/it-IT/electron.json +39 -0
- package/locales/it-IT/setting.json +1 -0
- package/locales/ja-JP/electron.json +39 -0
- package/locales/ja-JP/setting.json +1 -0
- package/locales/ko-KR/electron.json +39 -0
- package/locales/ko-KR/setting.json +1 -0
- package/locales/nl-NL/electron.json +39 -0
- package/locales/nl-NL/setting.json +1 -0
- package/locales/pl-PL/electron.json +39 -0
- package/locales/pl-PL/setting.json +1 -0
- package/locales/pt-BR/electron.json +39 -0
- package/locales/pt-BR/setting.json +1 -0
- package/locales/ru-RU/electron.json +39 -0
- package/locales/ru-RU/setting.json +1 -0
- package/locales/tr-TR/electron.json +39 -0
- package/locales/tr-TR/setting.json +1 -0
- package/locales/vi-VN/electron.json +39 -0
- package/locales/vi-VN/setting.json +1 -0
- package/locales/zh-CN/electron.json +39 -0
- package/locales/zh-CN/setting.json +1 -0
- package/locales/zh-TW/electron.json +39 -0
- package/locales/zh-TW/setting.json +1 -0
- package/package.json +3 -3
- package/packages/electron-client-ipc/src/events/index.ts +3 -1
- package/packages/electron-client-ipc/src/events/settings.ts +12 -0
- package/packages/electron-client-ipc/src/types/index.ts +1 -0
- package/packages/electron-client-ipc/src/types/proxy.ts +12 -0
- package/src/app/[variants]/(main)/settings/hooks/useCategory.tsx +11 -1
- package/src/app/[variants]/(main)/settings/proxy/features/ProxyForm.tsx +369 -0
- package/src/app/[variants]/(main)/settings/proxy/index.tsx +22 -0
- package/src/app/[variants]/(main)/settings/proxy/page.tsx +28 -0
- package/src/config/aiModels/google.ts +15 -26
- package/src/config/aiModels/groq.ts +0 -16
- package/src/config/aiModels/hunyuan.ts +79 -1
- package/src/config/aiModels/novita.ts +50 -56
- package/src/config/aiModels/qwen.ts +10 -32
- package/src/config/aiModels/siliconcloud.ts +111 -86
- package/src/config/aiModels/zhipu.ts +74 -12
- package/src/config/modelProviders/zhipu.ts +1 -2
- package/src/libs/model-runtime/hunyuan/index.ts +9 -1
- package/src/locales/default/electron.ts +39 -0
- package/src/locales/default/setting.ts +1 -0
- package/src/services/electron/settings.ts +33 -0
- package/src/store/electron/actions/settings.ts +55 -0
- package/src/store/electron/initialState.ts +12 -1
- package/src/store/electron/selectors/__tests__/desktopState.test.ts +3 -1
- package/src/store/electron/store.ts +4 -1
- package/src/store/global/initialState.ts +1 -0
- package/apps/desktop/scripts/pglite-server.ts +0 -14
@@ -2,6 +2,33 @@ import { AIChatModelCard } from '@/types/aiModel';
|
|
2
2
|
|
3
3
|
// https://novita.ai/pricing
|
4
4
|
const novitaChatModels: AIChatModelCard[] = [
|
5
|
+
{
|
6
|
+
abilities: {
|
7
|
+
reasoning: true,
|
8
|
+
},
|
9
|
+
contextWindowTokens: 131_072,
|
10
|
+
displayName: 'Kimi K2 Instruct',
|
11
|
+
id: 'moonshotai/kimi-k2-instruct',
|
12
|
+
pricing: {
|
13
|
+
input: 0.57,
|
14
|
+
output: 2.3,
|
15
|
+
},
|
16
|
+
type: 'chat',
|
17
|
+
},
|
18
|
+
{
|
19
|
+
abilities: {
|
20
|
+
reasoning: true,
|
21
|
+
vision: true,
|
22
|
+
},
|
23
|
+
contextWindowTokens: 65_536,
|
24
|
+
displayName: 'GLM 4.1V 9B Thinking',
|
25
|
+
id: 'thudm/glm-4.1v-9b-thinking',
|
26
|
+
pricing: {
|
27
|
+
input: 0.035,
|
28
|
+
output: 0.138,
|
29
|
+
},
|
30
|
+
type: 'chat',
|
31
|
+
},
|
5
32
|
{
|
6
33
|
contextWindowTokens: 120_000,
|
7
34
|
displayName: 'ERNIE 4.5 0.3B',
|
@@ -23,12 +50,15 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
23
50
|
type: 'chat',
|
24
51
|
},
|
25
52
|
{
|
53
|
+
abilities: {
|
54
|
+
reasoning: true,
|
55
|
+
},
|
26
56
|
contextWindowTokens: 123_000,
|
27
57
|
displayName: 'ERNIE 4.5 300B A47B Paddle',
|
28
58
|
id: 'baidu/ernie-4.5-300b-a47b-paddle',
|
29
59
|
pricing: {
|
30
|
-
input: 0.
|
31
|
-
output: 1,
|
60
|
+
input: 0.28,
|
61
|
+
output: 1.1,
|
32
62
|
},
|
33
63
|
type: 'chat',
|
34
64
|
},
|
@@ -48,6 +78,7 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
48
78
|
},
|
49
79
|
{
|
50
80
|
abilities: {
|
81
|
+
functionCall: true,
|
51
82
|
reasoning: true,
|
52
83
|
vision: true,
|
53
84
|
},
|
@@ -62,9 +93,10 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
62
93
|
},
|
63
94
|
{
|
64
95
|
abilities: {
|
96
|
+
functionCall: true,
|
65
97
|
reasoning: true,
|
66
98
|
},
|
67
|
-
contextWindowTokens:
|
99
|
+
contextWindowTokens: 1_000_000,
|
68
100
|
displayName: 'MiniMax M1 80K',
|
69
101
|
id: 'minimaxai/minimax-m1-80k',
|
70
102
|
pricing: {
|
@@ -220,6 +252,17 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
220
252
|
},
|
221
253
|
type: 'chat',
|
222
254
|
},
|
255
|
+
{
|
256
|
+
contextWindowTokens: 32_768,
|
257
|
+
description: 'Gemma 3 1B 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。',
|
258
|
+
displayName: 'Gemma 3 1B',
|
259
|
+
id: 'google/gemma-3-1b-it',
|
260
|
+
pricing: {
|
261
|
+
input: 0,
|
262
|
+
output: 0,
|
263
|
+
},
|
264
|
+
type: 'chat',
|
265
|
+
},
|
223
266
|
{
|
224
267
|
contextWindowTokens: 60_288,
|
225
268
|
description: 'Mistral Nemo 是多语言支持和高性能编程的7.3B参数模型。',
|
@@ -318,8 +361,8 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
318
361
|
enabled: true,
|
319
362
|
id: 'deepseek/deepseek-v3-0324',
|
320
363
|
pricing: {
|
321
|
-
input: 0.
|
322
|
-
output: 1.
|
364
|
+
input: 0.28,
|
365
|
+
output: 1.14,
|
323
366
|
},
|
324
367
|
type: 'chat',
|
325
368
|
},
|
@@ -527,55 +570,19 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
527
570
|
},
|
528
571
|
type: 'chat',
|
529
572
|
},
|
530
|
-
{
|
531
|
-
contextWindowTokens: 32_000,
|
532
|
-
displayName: 'Qwen2.5 7B Instruct',
|
533
|
-
id: 'qwen/qwen2.5-7b-instruct',
|
534
|
-
pricing: {
|
535
|
-
input: 0,
|
536
|
-
output: 0,
|
537
|
-
},
|
538
|
-
type: 'chat',
|
539
|
-
},
|
540
|
-
{
|
541
|
-
abilities: {
|
542
|
-
functionCall: true,
|
543
|
-
},
|
544
|
-
contextWindowTokens: 32_000,
|
545
|
-
displayName: 'GLM 4 9B 0414',
|
546
|
-
id: 'thudm/glm-4-9b-0414',
|
547
|
-
pricing: {
|
548
|
-
input: 0,
|
549
|
-
output: 0,
|
550
|
-
},
|
551
|
-
type: 'chat',
|
552
|
-
},
|
553
573
|
{
|
554
574
|
abilities: {
|
555
575
|
functionCall: true,
|
556
576
|
},
|
557
577
|
contextWindowTokens: 32_000,
|
558
|
-
displayName: '
|
559
|
-
id: '
|
578
|
+
displayName: 'Qwen2.5 7B Instruct',
|
579
|
+
id: 'qwen/qwen2.5-7b-instruct',
|
560
580
|
pricing: {
|
561
581
|
input: 0,
|
562
582
|
output: 0,
|
563
583
|
},
|
564
584
|
type: 'chat',
|
565
585
|
},
|
566
|
-
{
|
567
|
-
abilities: {
|
568
|
-
functionCall: true,
|
569
|
-
},
|
570
|
-
contextWindowTokens: 32_000,
|
571
|
-
displayName: 'GLM Z1 32B 0414',
|
572
|
-
id: 'thudm/glm-z1-32b-0414',
|
573
|
-
pricing: {
|
574
|
-
input: 0.24,
|
575
|
-
output: 0.24,
|
576
|
-
},
|
577
|
-
type: 'chat',
|
578
|
-
},
|
579
586
|
{
|
580
587
|
abilities: {
|
581
588
|
functionCall: true,
|
@@ -589,19 +596,6 @@ const novitaChatModels: AIChatModelCard[] = [
|
|
589
596
|
},
|
590
597
|
type: 'chat',
|
591
598
|
},
|
592
|
-
{
|
593
|
-
abilities: {
|
594
|
-
functionCall: true,
|
595
|
-
},
|
596
|
-
contextWindowTokens: 32_000,
|
597
|
-
displayName: 'GLM Z1 Rumination 32B 0414',
|
598
|
-
id: 'thudm/glm-z1-rumination-32b-0414',
|
599
|
-
pricing: {
|
600
|
-
input: 0.24,
|
601
|
-
output: 0.24,
|
602
|
-
},
|
603
|
-
type: 'chat',
|
604
|
-
},
|
605
599
|
];
|
606
600
|
|
607
601
|
export const allModels = [...novitaChatModels];
|
@@ -18,8 +18,8 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
18
18
|
organization: 'Qwen',
|
19
19
|
pricing: {
|
20
20
|
currency: 'CNY',
|
21
|
-
input:
|
22
|
-
output:
|
21
|
+
input: 2,
|
22
|
+
output: 20, // Thinking mode pricing
|
23
23
|
},
|
24
24
|
releasedAt: '2025-04-28',
|
25
25
|
settings: {
|
@@ -36,7 +36,6 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
36
36
|
description:
|
37
37
|
'Qwen3是一款能力大幅提升的新一代通义千问大模型,在推理、通用、Agent和多语言等多个核心能力上均达到业界领先水平,并支持思考模式切换。',
|
38
38
|
displayName: 'Qwen3 32B',
|
39
|
-
enabled: true,
|
40
39
|
id: 'qwen3-32b',
|
41
40
|
maxOutput: 8192,
|
42
41
|
organization: 'Qwen',
|
@@ -66,8 +65,8 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
66
65
|
organization: 'Qwen',
|
67
66
|
pricing: {
|
68
67
|
currency: 'CNY',
|
69
|
-
input:
|
70
|
-
output:
|
68
|
+
input: 0.75,
|
69
|
+
output: 7.5, // Thinking mode pricing
|
71
70
|
},
|
72
71
|
releasedAt: '2025-04-28',
|
73
72
|
settings: {
|
@@ -203,7 +202,6 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
203
202
|
description:
|
204
203
|
'基于 Qwen2.5 模型训练的 QwQ 推理模型,通过强化学习大幅度提升了模型推理能力。模型数学代码等核心指标(AIME 24/25、LiveCodeBench)以及部分通用指标(IFEval、LiveBench等)达到DeepSeek-R1 满血版水平。',
|
205
204
|
displayName: 'QwQ Plus',
|
206
|
-
enabled: true,
|
207
205
|
id: 'qwq-plus',
|
208
206
|
maxOutput: 8192,
|
209
207
|
organization: 'Qwen',
|
@@ -237,7 +235,7 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
237
235
|
pricing: {
|
238
236
|
currency: 'CNY',
|
239
237
|
input: 0.3,
|
240
|
-
output:
|
238
|
+
output: 3, // Thinking mode pricing
|
241
239
|
},
|
242
240
|
releasedAt: '2025-04-28',
|
243
241
|
settings: {
|
@@ -265,7 +263,7 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
265
263
|
pricing: {
|
266
264
|
currency: 'CNY',
|
267
265
|
input: 0.8,
|
268
|
-
output:
|
266
|
+
output: 8, // Thinking mode pricing
|
269
267
|
},
|
270
268
|
releasedAt: '2025-04-28',
|
271
269
|
settings: {
|
@@ -311,7 +309,6 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
311
309
|
description:
|
312
310
|
'通义千问超大规模语言模型,支持长文本上下文,以及基于长文档、多文档等多个场景的对话功能。',
|
313
311
|
displayName: 'Qwen Long',
|
314
|
-
enabled: true,
|
315
312
|
id: 'qwen-long',
|
316
313
|
maxOutput: 8192,
|
317
314
|
organization: 'Qwen',
|
@@ -327,7 +324,7 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
327
324
|
vision: true,
|
328
325
|
},
|
329
326
|
config: {
|
330
|
-
deploymentName: 'qwen-omni-turbo-latest',
|
327
|
+
deploymentName: 'qwen-omni-turbo-latest', // expired on 2025-08-13
|
331
328
|
},
|
332
329
|
contextWindowTokens: 32_768,
|
333
330
|
description:
|
@@ -357,7 +354,7 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
357
354
|
organization: 'Qwen',
|
358
355
|
pricing: {
|
359
356
|
currency: 'CNY',
|
360
|
-
input:
|
357
|
+
input: 2, // use image input price
|
361
358
|
output: 6,
|
362
359
|
},
|
363
360
|
type: 'chat',
|
@@ -818,12 +815,12 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
818
815
|
abilities: {
|
819
816
|
reasoning: true,
|
820
817
|
},
|
821
|
-
contextWindowTokens:
|
818
|
+
contextWindowTokens: 131_072,
|
822
819
|
description:
|
823
820
|
'685B 满血版模型,2025年5月28日发布。DeepSeek-R1 在后训练阶段大规模使用了强化学习技术,在仅有极少标注数据的情况下,极大提升了模型推理能力。在数学、代码、自然语言推理等任务上,性能较高,能力较强。',
|
824
821
|
displayName: 'DeepSeek R1 0528',
|
825
822
|
id: 'deepseek-r1-0528',
|
826
|
-
maxOutput:
|
823
|
+
maxOutput: 16_384,
|
827
824
|
organization: 'DeepSeek',
|
828
825
|
pricing: {
|
829
826
|
currency: 'CNY',
|
@@ -833,25 +830,6 @@ const qwenChatModels: AIChatModelCard[] = [
|
|
833
830
|
releasedAt: '2025-05-28',
|
834
831
|
type: 'chat',
|
835
832
|
},
|
836
|
-
{
|
837
|
-
abilities: {
|
838
|
-
reasoning: true,
|
839
|
-
},
|
840
|
-
contextWindowTokens: 65_536,
|
841
|
-
description:
|
842
|
-
'671B 满血版模型,2025年1月20日发布。DeepSeek-R1 在后训练阶段大规模使用了强化学习技术,在仅有极少标注数据的情况下,极大提升了模型推理能力。在数学、代码、自然语言推理等任务上,性能较高,能力较强。',
|
843
|
-
displayName: 'DeepSeek R1',
|
844
|
-
id: 'deepseek-r1',
|
845
|
-
maxOutput: 8192,
|
846
|
-
organization: 'DeepSeek',
|
847
|
-
pricing: {
|
848
|
-
currency: 'CNY',
|
849
|
-
input: 4,
|
850
|
-
output: 16,
|
851
|
-
},
|
852
|
-
releasedAt: '2025-01-27',
|
853
|
-
type: 'chat',
|
854
|
-
},
|
855
833
|
{
|
856
834
|
contextWindowTokens: 65_536,
|
857
835
|
description:
|
@@ -2,6 +2,37 @@ import { AIChatModelCard } from '@/types/aiModel';
|
|
2
2
|
|
3
3
|
// https://siliconflow.cn/zh-cn/models
|
4
4
|
const siliconcloudChatModels: AIChatModelCard[] = [
|
5
|
+
{
|
6
|
+
abilities: {
|
7
|
+
reasoning: true,
|
8
|
+
},
|
9
|
+
contextWindowTokens: 131_072,
|
10
|
+
description:
|
11
|
+
'Pangu-Pro-MoE 72B-A16B 是一款 720 亿参数、激活 160 亿参的稀疏大语言模型,它基于分组混合专家(MoGE)架构,它在专家选择阶段对专家进行分组,并约束 token 在每个组内激活等量专家,从而实现专家负载均衡,显著提升模型在昇腾平台的部署效率。',
|
12
|
+
displayName: 'Pangu Pro MoE 72B A16B',
|
13
|
+
id: 'ascend-tribe/pangu-pro-moe',
|
14
|
+
pricing: {
|
15
|
+
currency: 'CNY',
|
16
|
+
input: 1,
|
17
|
+
output: 4,
|
18
|
+
},
|
19
|
+
releasedAt: '2025-06-17',
|
20
|
+
type: 'chat',
|
21
|
+
},
|
22
|
+
{
|
23
|
+
contextWindowTokens: 131_072,
|
24
|
+
description:
|
25
|
+
'ERNIE-4.5-300B-A47B 是由百度公司开发的一款基于混合专家(MoE)架构的大语言模型。该模型总参数量为 3000 亿,但在推理时每个 token 仅激活 470 亿参数,从而在保证强大性能的同时兼顾了计算效率。作为 ERNIE 4.5 系列的核心模型之一,在文本理解、生成、推理和编程等任务上展现出卓越的能力。该模型采用了一种创新的多模态异构 MoE 预训练方法,通过文本与视觉模态的联合训练,有效提升了模型的综合能力,尤其在指令遵循和世界知识记忆方面效果突出。',
|
26
|
+
displayName: 'ERNIE 4.5 300B A47B',
|
27
|
+
id: 'baidu/ERNIE-4.5-300B-A47B',
|
28
|
+
pricing: {
|
29
|
+
currency: 'CNY',
|
30
|
+
input: 2,
|
31
|
+
output: 8,
|
32
|
+
},
|
33
|
+
releasedAt: '2025-06-30',
|
34
|
+
type: 'chat',
|
35
|
+
},
|
5
36
|
{
|
6
37
|
abilities: {
|
7
38
|
reasoning: true,
|
@@ -166,7 +197,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
166
197
|
contextWindowTokens: 131_072,
|
167
198
|
description:
|
168
199
|
'Qwen3是一款能力大幅提升的新一代通义千问大模型,在推理、通用、Agent和多语言等多个核心能力上均达到业界领先水平,并支持思考模式切换。',
|
169
|
-
displayName: 'Qwen3 8B',
|
200
|
+
displayName: 'Qwen3 8B (Free)',
|
170
201
|
enabled: true,
|
171
202
|
id: 'Qwen/Qwen3-8B',
|
172
203
|
organization: 'Qwen',
|
@@ -181,6 +212,43 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
181
212
|
},
|
182
213
|
type: 'chat',
|
183
214
|
},
|
215
|
+
{
|
216
|
+
abilities: {
|
217
|
+
reasoning: true,
|
218
|
+
vision: true,
|
219
|
+
},
|
220
|
+
contextWindowTokens: 65_536,
|
221
|
+
description:
|
222
|
+
'GLM-4.1V-9B-Thinking 是由智谱 AI 和清华大学 KEG 实验室联合发布的一款开源视觉语言模型(VLM),专为处理复杂的多模态认知任务而设计。该模型基于 GLM-4-9B-0414 基础模型,通过引入“思维链”(Chain-of-Thought)推理机制和采用强化学习策略,显著提升了其跨模态的推理能力和稳定性。',
|
223
|
+
displayName: 'GLM-4.1V 9B Thinking (Free)',
|
224
|
+
enabled: true,
|
225
|
+
id: 'THUDM/GLM-4.1V-9B-Thinking',
|
226
|
+
pricing: {
|
227
|
+
currency: 'CNY',
|
228
|
+
input: 0,
|
229
|
+
output: 0,
|
230
|
+
},
|
231
|
+
releasedAt: '2025-07-02',
|
232
|
+
type: 'chat',
|
233
|
+
},
|
234
|
+
{
|
235
|
+
abilities: {
|
236
|
+
reasoning: true,
|
237
|
+
vision: true,
|
238
|
+
},
|
239
|
+
contextWindowTokens: 65_536,
|
240
|
+
description:
|
241
|
+
'GLM-4.1V-9B-Thinking 是由智谱 AI 和清华大学 KEG 实验室联合发布的一款开源视觉语言模型(VLM),专为处理复杂的多模态认知任务而设计。该模型基于 GLM-4-9B-0414 基础模型,通过引入“思维链”(Chain-of-Thought)推理机制和采用强化学习策略,显著提升了其跨模态的推理能力和稳定性。',
|
242
|
+
displayName: 'GLM-4.1V 9B Thinking (Pro)',
|
243
|
+
id: 'Pro/THUDM/GLM-4.1V-9B-Thinking',
|
244
|
+
pricing: {
|
245
|
+
currency: 'CNY',
|
246
|
+
input: 0.25,
|
247
|
+
output: 1,
|
248
|
+
},
|
249
|
+
releasedAt: '2025-07-02',
|
250
|
+
type: 'chat',
|
251
|
+
},
|
184
252
|
{
|
185
253
|
abilities: {
|
186
254
|
reasoning: true,
|
@@ -195,6 +263,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
195
263
|
input: 1,
|
196
264
|
output: 4,
|
197
265
|
},
|
266
|
+
releasedAt: '2025-04-14',
|
198
267
|
type: 'chat',
|
199
268
|
},
|
200
269
|
{
|
@@ -212,6 +281,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
212
281
|
input: 1,
|
213
282
|
output: 4,
|
214
283
|
},
|
284
|
+
releasedAt: '2025-04-14',
|
215
285
|
type: 'chat',
|
216
286
|
},
|
217
287
|
{
|
@@ -222,13 +292,14 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
222
292
|
contextWindowTokens: 131_072,
|
223
293
|
description:
|
224
294
|
'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。',
|
225
|
-
displayName: 'GLM-Z1 9B 0414',
|
295
|
+
displayName: 'GLM-Z1 9B 0414 (Free)',
|
226
296
|
id: 'THUDM/GLM-Z1-9B-0414',
|
227
297
|
pricing: {
|
228
298
|
currency: 'CNY',
|
229
299
|
input: 0,
|
230
300
|
output: 0,
|
231
301
|
},
|
302
|
+
releasedAt: '2025-04-14',
|
232
303
|
type: 'chat',
|
233
304
|
},
|
234
305
|
{
|
@@ -245,6 +316,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
245
316
|
input: 1.89,
|
246
317
|
output: 1.89,
|
247
318
|
},
|
319
|
+
releasedAt: '2025-04-14',
|
248
320
|
type: 'chat',
|
249
321
|
},
|
250
322
|
{
|
@@ -254,14 +326,48 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
254
326
|
contextWindowTokens: 32_768,
|
255
327
|
description:
|
256
328
|
'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。',
|
257
|
-
displayName: 'GLM-4 9B 0414',
|
258
|
-
enabled: true,
|
329
|
+
displayName: 'GLM-4 9B 0414 (Free)',
|
259
330
|
id: 'THUDM/GLM-4-9B-0414',
|
260
331
|
pricing: {
|
261
332
|
currency: 'CNY',
|
262
333
|
input: 0,
|
263
334
|
output: 0,
|
264
335
|
},
|
336
|
+
releasedAt: '2025-04-14',
|
337
|
+
type: 'chat',
|
338
|
+
},
|
339
|
+
{
|
340
|
+
abilities: {
|
341
|
+
functionCall: true,
|
342
|
+
},
|
343
|
+
contextWindowTokens: 131_072,
|
344
|
+
description:
|
345
|
+
'GLM-4-9B-Chat 是智谱 AI 推出的 GLM-4 系列预训练模型中的开源版本。该模型在语义、数学、推理、代码和知识等多个方面表现出色。除了支持多轮对话外,GLM-4-9B-Chat 还具备网页浏览、代码执行、自定义工具调用(Function Call)和长文本推理等高级功能。模型支持 26 种语言,包括中文、英文、日语、韩语和德语等。在多项基准测试中,GLM-4-9B-Chat 展现了优秀的性能,如 AlignBench-v2、MT-Bench、MMLU 和 C-Eval 等。该模型支持最大 128K 的上下文长度,适用于学术研究和商业应用',
|
346
|
+
displayName: 'GLM-4 9B Chat (Free)',
|
347
|
+
id: 'THUDM/glm-4-9b-chat',
|
348
|
+
pricing: {
|
349
|
+
currency: 'CNY',
|
350
|
+
input: 0,
|
351
|
+
output: 0,
|
352
|
+
},
|
353
|
+
releasedAt: '2024-06-04',
|
354
|
+
type: 'chat',
|
355
|
+
},
|
356
|
+
{
|
357
|
+
abilities: {
|
358
|
+
functionCall: true,
|
359
|
+
},
|
360
|
+
contextWindowTokens: 131_072,
|
361
|
+
description:
|
362
|
+
'GLM-4-9B-Chat 是智谱 AI 推出的 GLM-4 系列预训练模型中的开源版本。该模型在语义、数学、推理、代码和知识等多个方面表现出色。除了支持多轮对话外,GLM-4-9B-Chat 还具备网页浏览、代码执行、自定义工具调用(Function Call)和长文本推理等高级功能。模型支持 26 种语言,包括中文、英文、日语、韩语和德语等。在多项基准测试中,GLM-4-9B-Chat 展现了优秀的性能,如 AlignBench-v2、MT-Bench、MMLU 和 C-Eval 等。该模型支持最大 128K 的上下文长度,适用于学术研究和商业应用',
|
363
|
+
displayName: 'GLM-4 9B Chat (Pro)',
|
364
|
+
id: 'Pro/THUDM/glm-4-9b-chat',
|
365
|
+
pricing: {
|
366
|
+
currency: 'CNY',
|
367
|
+
input: 0.6,
|
368
|
+
output: 0.6,
|
369
|
+
},
|
370
|
+
releasedAt: '2024-06-04',
|
265
371
|
type: 'chat',
|
266
372
|
},
|
267
373
|
{
|
@@ -271,7 +377,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
271
377
|
contextWindowTokens: 131_072,
|
272
378
|
description:
|
273
379
|
'DeepSeek-R1-0528-Qwen3-8B 是通过从 DeepSeek-R1-0528 模型蒸馏思维链到 Qwen3 8B Base 获得的模型。该模型在开源模型中达到了最先进(SOTA)的性能,在 AIME 2024 测试中超越了 Qwen3 8B 10%,并达到了 Qwen3-235B-thinking 的性能水平。该模型在数学推理、编程和通用逻辑等多个基准测试中表现出色,其架构与 Qwen3-8B 相同,但共享 DeepSeek-R1-0528 的分词器配置。',
|
274
|
-
displayName: 'DeepSeek R1 0528 Qwen3 8B',
|
380
|
+
displayName: 'DeepSeek R1 0528 Qwen3 8B (Free)',
|
275
381
|
enabled: true,
|
276
382
|
id: 'deepseek-ai/DeepSeek-R1-0528-Qwen3-8B',
|
277
383
|
pricing: {
|
@@ -331,23 +437,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
331
437
|
},
|
332
438
|
type: 'chat',
|
333
439
|
},
|
334
|
-
{
|
335
|
-
abilities: {
|
336
|
-
functionCall: true,
|
337
|
-
reasoning: true,
|
338
|
-
},
|
339
|
-
contextWindowTokens: 98_304,
|
340
|
-
description:
|
341
|
-
'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。',
|
342
|
-
displayName: 'DeepSeek R1 0120 (Pro)',
|
343
|
-
id: 'Pro/deepseek-ai/DeepSeek-R1-0120',
|
344
|
-
pricing: {
|
345
|
-
currency: 'CNY',
|
346
|
-
input: 4,
|
347
|
-
output: 16,
|
348
|
-
},
|
349
|
-
type: 'chat',
|
350
|
-
},
|
351
440
|
{
|
352
441
|
abilities: {
|
353
442
|
functionCall: true,
|
@@ -364,22 +453,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
364
453
|
},
|
365
454
|
type: 'chat',
|
366
455
|
},
|
367
|
-
{
|
368
|
-
abilities: {
|
369
|
-
functionCall: true,
|
370
|
-
},
|
371
|
-
contextWindowTokens: 65_536,
|
372
|
-
description:
|
373
|
-
'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。',
|
374
|
-
displayName: 'DeepSeek V3 1226 (Pro)',
|
375
|
-
id: 'Pro/deepseek-ai/DeepSeek-V3-1226',
|
376
|
-
pricing: {
|
377
|
-
currency: 'CNY',
|
378
|
-
input: 2,
|
379
|
-
output: 8,
|
380
|
-
},
|
381
|
-
type: 'chat',
|
382
|
-
},
|
383
456
|
{
|
384
457
|
abilities: {
|
385
458
|
functionCall: true,
|
@@ -514,22 +587,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
514
587
|
},
|
515
588
|
type: 'chat',
|
516
589
|
},
|
517
|
-
{
|
518
|
-
abilities: {
|
519
|
-
reasoning: true,
|
520
|
-
},
|
521
|
-
contextWindowTokens: 32_768,
|
522
|
-
description:
|
523
|
-
'QwQ-32B-Preview 是 Qwen 最新的实验性研究模型,专注于提升AI推理能力。通过探索语言混合、递归推理等复杂机制,主要优势包括强大的推理分析能力、数学和编程能力。与此同时,也存在语言切换问题、推理循环、安全性考虑、其他能力方面的差异。',
|
524
|
-
displayName: 'QwQ 32B Preview',
|
525
|
-
id: 'Qwen/QwQ-32B-Preview',
|
526
|
-
pricing: {
|
527
|
-
currency: 'CNY',
|
528
|
-
input: 1.26,
|
529
|
-
output: 1.26,
|
530
|
-
},
|
531
|
-
type: 'chat',
|
532
|
-
},
|
533
590
|
{
|
534
591
|
abilities: {
|
535
592
|
functionCall: true,
|
@@ -771,38 +828,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
|
|
771
828
|
},
|
772
829
|
type: 'chat',
|
773
830
|
},
|
774
|
-
{
|
775
|
-
abilities: {
|
776
|
-
functionCall: true,
|
777
|
-
},
|
778
|
-
contextWindowTokens: 131_072,
|
779
|
-
description:
|
780
|
-
'GLM-4-9B-Chat 是智谱 AI 推出的 GLM-4 系列预训练模型中的开源版本。该模型在语义、数学、推理、代码和知识等多个方面表现出色。除了支持多轮对话外,GLM-4-9B-Chat 还具备网页浏览、代码执行、自定义工具调用(Function Call)和长文本推理等高级功能。模型支持 26 种语言,包括中文、英文、日语、韩语和德语等。在多项基准测试中,GLM-4-9B-Chat 展现了优秀的性能,如 AlignBench-v2、MT-Bench、MMLU 和 C-Eval 等。该模型支持最大 128K 的上下文长度,适用于学术研究和商业应用',
|
781
|
-
displayName: 'GLM-4 9B Chat (Free)',
|
782
|
-
id: 'THUDM/glm-4-9b-chat',
|
783
|
-
pricing: {
|
784
|
-
currency: 'CNY',
|
785
|
-
input: 0,
|
786
|
-
output: 0,
|
787
|
-
},
|
788
|
-
type: 'chat',
|
789
|
-
},
|
790
|
-
{
|
791
|
-
abilities: {
|
792
|
-
functionCall: true,
|
793
|
-
},
|
794
|
-
contextWindowTokens: 131_072,
|
795
|
-
description:
|
796
|
-
'GLM-4-9B-Chat 是智谱 AI 推出的 GLM-4 系列预训练模型中的开源版本。该模型在语义、数学、推理、代码和知识等多个方面表现出色。除了支持多轮对话外,GLM-4-9B-Chat 还具备网页浏览、代码执行、自定义工具调用(Function Call)和长文本推理等高级功能。模型支持 26 种语言,包括中文、英文、日语、韩语和德语等。在多项基准测试中,GLM-4-9B-Chat 展现了优秀的性能,如 AlignBench-v2、MT-Bench、MMLU 和 C-Eval 等。该模型支持最大 128K 的上下文长度,适用于学术研究和商业应用',
|
797
|
-
displayName: 'GLM-4 9B Chat (Pro)',
|
798
|
-
id: 'Pro/THUDM/glm-4-9b-chat',
|
799
|
-
pricing: {
|
800
|
-
currency: 'CNY',
|
801
|
-
input: 0.6,
|
802
|
-
output: 0.6,
|
803
|
-
},
|
804
|
-
type: 'chat',
|
805
|
-
},
|
806
831
|
];
|
807
832
|
|
808
833
|
export const allModels = [...siliconcloudChatModels];
|