AIHOT LEADERBOARD
AI 模型排行榜
只用第三方评测机构给所有模型统一跑的成绩,换算到同一条能力刻度上:编程 40%、科研 20%、专业办公 20%、知识问答 15%、真人盲选 5%。
35 项评测 · 7 家机构 · 10/10 08:06 更新
数据同步自 AIHOT 官方模型榜 ↗,沿用官方排名与评分。
综合榜 30 个模型
| 名次 | 模型 | |||
|---|---|---|---|---|
| 01 | Claude Opus 5.5Anthropic | 2026-09-22 | 33 项评测覆盖 96% | 73.9 |
| 02 | Gemini 4 ArgonGoogle | 2026-09-30 | 21 项评测覆盖 68% | 73.5 |
| 03 | Claude Fable 5.1Anthropic | 2026-09-01 | 35 项评测覆盖 100% | 72.2 |
| 04 | Claude Sonnet 5.5Anthropic | 2026-09-28 | 33 项评测覆盖 100% | 71.4 |
| 05 | GPT-6 AstraOpenAI | 2026-09-03 | 35 项评测覆盖 100% | 71.3 |
| 06 | Claude Fable 5Anthropic | 2026-06-09 | 30 项评测覆盖 89% | 69.9 |
| 07 | GPT-6.1 SolOpenAI | 2026-09-29 | 34 项评测覆盖 100% | 69.0 |
| 08 | Claude Opus 5Anthropic | 2026-07-24 | 35 项评测覆盖 100% | 68.9 |
| 09 | GPT-5.6 SolOpenAI | 2026-07-09 | 33 项评测覆盖 93% | 65.9 |
| 10 | GPT-6 SolOpenAI | 2026-09-22 | 32 项评测覆盖 100% | 65.6 |
| 11 | Muse Spark 1.3Meta | 2026-09-02 | 31 项评测覆盖 93% | 65.6 |
| 12 | Gemini 3.8 FlashGoogle | 2026-09-02 | 33 项评测覆盖 96% | 64.4 |
| 13 | Kimi K3Moonshot AI开源权重 ↗ | 2026-07-16 | 32 项评测覆盖 93% | 63.2 |
| 14 | Claude Opus 4.8Anthropic | 2026-05-28 | 30 项评测覆盖 89% | 63.0 |
| 15 | MiMo-V2.6-ProXiaomi | 2026-09-21 | 23 项评测覆盖 68% | 62.9 |
| 16 | Gemini 3.7 FlashGoogle | 2026-08-13 | 31 项评测覆盖 93% | 62.8 |
| 17 | GLM-5.3Z.ai开源权重 ↗ | 2026-08-18 | 34 项评测覆盖 100% | 62.4 |
| 18 | GPT-5.5OpenAI | 2026-04-23 | 28 项评测覆盖 86% | 62.3 |
| 19 | Claude Opus 4.7Anthropic | 2026-04-16 | 19 项评测覆盖 64% | 62.1 |
| 20 | GPT-5.6 TerraOpenAI | 2026-07-09 | 33 项评测覆盖 96% | 62.1 |
| 21 | Grok 4.7xAI | 2026-09-21 | 31 项评测覆盖 96% | 61.6 |
| 22 | Grok 4.6xAI | 2026-08-12 | 35 项评测覆盖 100% | 61.2 |
| 23 | Muse Spark 1.2Meta | 2026-08-05 | 26 项评测覆盖 82% | 61.1 |
| 24 | Step 5 PreviewStepFun | 2026-09-18 | 20 项评测覆盖 57% | 60.5 |
| 25 | DeepSeek V4.1 FlashDeepSeek开源权重 ↗ | 2026-09-10 | 32 项评测覆盖 100% | 60.4 |
| 26 | Claude Haiku 5.5Anthropic | 2026-10-07 | 31 项评测覆盖 93% | 60.1 |
| 27 | DeepSeek V4 Pro 0813DeepSeek开源权重 ↗ | 2026-08-13 | 33 项评测覆盖 96% | 58.8 |
| 28 | Claude Sonnet 5Anthropic | 2026-06-30 | 32 项评测覆盖 93% | 58.7 |
| 29 | Qwen3.8 MaxAlibaba | 2026-07-19 | 33 项评测覆盖 100% | 58.6 |
| 30 | Muse Spark 1.1Meta | 2026-07-09 | 21 项评测覆盖 71% | 58.5 |
AIHOT 评分把各家评测换算到同一条能力刻度上:50 分是参照组模型的平均水平,每 15 分约是一个标准差。评分不是正确率。 价格为 API 参考报价,原始出处见模型详情。