只用第三方评测机构实测的成绩,厂商自己公布的分数一概不计。各项成绩折算成同一种分数后按领域加权:编程 38%、创意生成 29%、专业办公 19%、知识问答 14%。
AI 浪潮席卷科研领域,新能力、新工具迭代极快,推荐各位科研伙伴持续学习 AI 相关知识,探索大模型在本方向的落地方式,借助 AI 赋能科研创新。
37 项评测·54 个模型·2026-11-11 更新
排序
| 名次 | 模型 | 上线日期 | 评测覆盖 | 缓存价格 | 输入价格 | 输出价格 | 综合评分 |
|---|---|---|---|---|---|---|---|
| 01 | Claude Fable 5.2 Anthropic | 2026-11-09 | 23 项测评 覆盖 62% | ¥1.33 | ¥53.36 | ¥266.80 | 65.9 |
| 02 | Claude Opus 5.5 Anthropic | 2026-09-22 | 27 项测评 覆盖 73% | ¥1.33 | ¥26.68 | ¥133.40 | 64.1 |
| 03 | GPT-6 Astra OpenAI | 2026-09-03 | 30 项测评 覆盖 81% | ¥6.67 | ¥66.70 | ¥333.50 | 63.8 |
| 04 | Gemini 4 Carbon Google | 2026-10-28 | 15 项测评 覆盖 41% | ¥1.33 | ¥26.68 | ¥133.40 | 62.6 |
| 05 | Claude Sonnet 5.5 Anthropic | 2026-09-28 | 25 项测评 覆盖 68% | ¥0.67 | ¥13.34 | ¥66.70 | 62 |
| 06 | GPT-6.1 Sol OpenAI | 2026-09-29 | 26 项测评 覆盖 70% | ¥0.67 | ¥13.34 | ¥66.70 | 60.5 |
| 07 | Gemini 4 Argon Google | 2026-09-30 | 25 项测评 覆盖 68% | ¥1.33 | ¥26.68 | ¥133.40 | 58.4 |
| 08 | Muse Wonder Full Meta | 2026-10-12 | 18 项测评 覆盖 49% | — | — | — | 56 |
| 09 | GPT-6.1 Terra OpenAI | 2026-10-26 | 20 项测评 覆盖 54% | ¥0.33 | ¥6.67 | ¥33.35 | 54.2 |
| 10 | Kimi K3.1 Pro Kimi开源国产 | 2026-10-23 | 22 项测评 覆盖 59% | ¥1.33 | ¥13.34 | ¥66.70 | 54.2 |
| 11 | Muse Wonder Meta | 2026-10-12 | 21 项测评 覆盖 57% | ¥1.33 | ¥26.68 | ¥133.40 | 52.5 |
| 12 | Qwen3.8 Max (0902) Alibaba国产 | 2026-08-03 | 29 项测评 覆盖 78% | ¥1.67 | ¥13.34 | ¥40.02 | 49.6 |
| 13 | Claude Haiku 5.5 Anthropic | 2026-10-07 | 25 项测评 覆盖 68% | ¥0.07 | ¥0.67 | ¥3.34 | 49 |
| 14 | GLM-5.3 Z AI开源国产 | 2026-08-18 | 28 项测评 覆盖 76% | ¥1.73 | ¥9.34 | ¥29.35 | 48.2 |
| 15 | Inkling 2 Small Thinking Machines | — | 23 项测评 覆盖 62% | ¥0.20 | ¥2 | ¥6.67 | 48.1 |
| 16 | Gemini 3.8 Flash Google | 2026-09-02 | 30 项测评 覆盖 81% | ¥0.50 | ¥5 | ¥25.01 | 48 |
| 17 | Grok 4.7 SpaceXAI | 2026-09-21 | 26 项测评 覆盖 70% | ¥3.34 | ¥13.34 | ¥40.02 | 47.3 |
| 18 | GLM-5.3-Flash Z AI开源国产 | 2026-08-26 | 28 项测评 覆盖 76% | ¥0.20 | ¥1 | ¥3.34 | 45 |
| 19 | MiMo-V2.6-Pro Xiaomi开源国产 | 2026-09-21 | 22 项测评 覆盖 59% | ¥0.02 | ¥2.90 | ¥5.80 | 44.3 |
| 20 | DeepSeek V4.1 Flash 1017 DeepSeek开源国产 | 2026-10-17 | 20 项测评 覆盖 54% | ¥0.04 | ¥2 | ¥8 | 43.7 |
| 21 | GPT-6 Luna OpenAI | 2026-09-22 | 26 项测评 覆盖 70% | ¥0.07 | ¥0.67 | ¥3.34 | 42.9 |
| 22 | Muse Spark Open Meta开源 | 2026-10-15 | 20 项测评 覆盖 54% | ¥1 | ¥8.34 | ¥28.35 | 42.8 |
| 23 | Step 5 Preview StepFun开源国产 | — | 25 项测评 覆盖 68% | ¥0.67 | ¥6.67 | ¥18.01 | 42.5 |
| 24 | Qwen3.8 27B Alibaba开源国产 | 2026-08-14 | 23 项测评 覆盖 62% | ¥0.67 | ¥3.34 | ¥20.01 | 41.8 |
| 25 | MiMo-V2.6-Flash Xiaomi开源国产 | 2026-09-21 | 19 项测评 覆盖 51% | ¥0.02 | ¥0.93 | ¥1.87 | 41.4 |
| 26 | Ling 3.1 Flash 1019 InclusionAI开源国产 | — | 16 项测评 覆盖 43% | ¥0.40 | ¥2 | ¥6 | 39.9 |
| 27 | DeepSeek V4 Pro 0813 DeepSeek开源国产 | 2026-08-13 | 20 项测评 覆盖 54% | ¥0.29 | ¥8.80 | ¥26.41 | 38.8 |
| 28 | MiniMax-M3.1-Flash MiniMax开源国产 | 2026-10-19 | 20 项测评 覆盖 54% | ¥0.10 | ¥1 | ¥4 | 38.6 |
| 29 | Muse Ember Meta | 2026-11-09 | 11 项测评 覆盖 30% | ¥0.03 | ¥0.67 | ¥3.34 | 38.6 |
| 30 | MiniMax-M3 MiniMax开源国产 | 2026-06-01 | 29 项测评 覆盖 78% | ¥0.40 | ¥2 | ¥8 | 37.7 |
| 31 | Agnes 2.5 Pro Theta Sapiens AI | 2026-11-07 | 16 项测评 覆盖 43% | ¥0.07 | ¥0.67 | ¥2 | 36.9 |
| 32 | Tencent Hy4 Tencent国产 | — | 10 项测评 覆盖 27% | ¥0.28 | ¥5.56 | ¥16.68 | 35.5 |
| 33 | K2 Horizon 375B A23B Institute of Foundation Models开源 | 2026-09-03 | 20 项测评 覆盖 54% | — | ¥0 | ¥0 | 34.9 |
| 34 | Apodex 1.1 Apodex | 2026-08-30 | 20 项测评 覆盖 54% | ¥0.20 | ¥2 | ¥20.01 | 33.5 |
| 35 | Beam 1.0 Reflection开源 | 2026-10-16 | 18 项测评 覆盖 49% | ¥0.27 | ¥2.67 | ¥8 | 31.6 |
| 36 | Gemini 3.5 Flash-Lite Google | 2026-07-21 | 22 项测评 覆盖 59% | ¥0.20 | ¥2 | ¥16.68 | 30.6 |
| 37 | — | 27 项测评 覆盖 73% | ¥0.80 | ¥4 | ¥18.01 | 30.6 | |
| 38 | Inkling Thinking Machines开源 | 2026-07-15 | 26 项测评 覆盖 70% | ¥1.13 | ¥6.67 | ¥27.01 | 30.2 |
| 39 | Inkling Small Thinking Machines开源 | 2026-07-30 | 24 项测评 覆盖 65% | ¥0.40 | ¥2 | ¥8 | 29.3 |
| 40 | Gemma 4 31B (Reasoning) Google开源 | 2026-04-02 | 24 项测评 覆盖 65% | — | ¥0 | ¥0 | 28.1 |
| 41 | Mistral Large 4 Mistral开源 | 2026-10-29 | 14 项测评 覆盖 38% | ¥0.47 | ¥4.67 | ¥14.01 | 27.5 |
| 42 | Mistral Medium 3.5 Mistral开源 | 2026-04-29 | 25 项测评 覆盖 68% | ¥1 | ¥10 | ¥50.03 | 27.1 |
| 43 | Muse Glimmer Meta开源 | 2026-08-10 | 22 项测评 覆盖 59% | ¥0.27 | ¥2 | ¥8 | 26.3 |
| 44 | Nemotron 3.5 Super VL NVIDIA开源 | 2026-10-15 | 17 项测评 覆盖 46% | ¥0.33 | ¥1.17 | ¥3.34 | 26.3 |
| 45 | Nanbeige4.3-33B Nanbeige开源国产 | 2026-10-25 | 10 项测评 覆盖 27% | — | — | — | 24.5 |
| 46 | K-EXAONE 2.0 LG AI Research开源 | 2026-08-12 | 18 项测评 覆盖 49% | — | — | — | 23.8 |
| 47 | Command A+ Cohere开源 | 2026-05-20 | 24 项测评 覆盖 65% | — | ¥0 | ¥0 | 23.7 |
| 48 | gpt-oss-120b OpenAI开源 | 2025-08-05 | 28 项测评 覆盖 76% | ¥0.67 | ¥1 | ¥4 | 22.3 |
| 49 | North Mini Code Cohere开源 | 2026-06-09 | 22 项测评 覆盖 59% | — | ¥0 | ¥0 | 19 |
| 50 | Nemotron 3.5 Lightning NVIDIA开源 | 2026-08-11 | 20 项测评 覆盖 54% | ¥0.27 | ¥0.40 | ¥1.33 | 18.9 |
| 51 | Mistral Small 4 (Reasoning) Mistral开源 | 2026-03-16 | 24 项测评 覆盖 65% | ¥0.10 | ¥1 | ¥4 | 18.4 |
| 52 | Granite 4.2 8B IBM开源 | 2026-08-25 | 18 项测评 覆盖 49% | ¥0.10 | ¥0.40 | ¥1.67 | 17 |
| 53 | Granite 4.2 3B IBM开源 | 2026-08-25 | 18 项测评 覆盖 49% | — | — | — | 13.6 |
| 54 | NVIDIA Nemotron 3 Nano 4B NVIDIA开源 | 2026-03-16 | 21 项测评 覆盖 57% | — | — | — | 9.6 |
价格单位:人民币 / 百万 tokens;「缓存价格」是命中缓存后的输入价。49 个模型的定价已核对,其余显示「待核」。点模型名看单页:备注、定价口径与来源链接。
分数是加权后的结果,口径写死在这里。原始成绩来自第三方评测机构,本站只做「按领域归一 + 加权」这一步,权重是上面那四个数,改动会记进更新日志。 评测覆盖是必须看的:「37 项里只测了 10 项」和「测满 30 项」不能直接比名次。 完整规则在怎么算的。



