AIHOT LEADERBOARD
AI 模型排行榜
事实准确、长文理解、语言与按要求作答、识破错误前提，看模型知道多少、答得多稳。
6 项评测 · 4 家机构 · 10/11 14:06 更新
数据同步自 AIHOT 官方模型榜 ↗，沿用官方排名与评分。
知识问答榜 30 个模型
| 名次 | 模型 | |||
|---|---|---|---|---|
| 01 | GPT-6 AstraOpenAI | 2026-09-03 | 6 项评测覆盖 100% | 74.7 |
| 02 | Claude Fable 5.1Anthropic | 2026-09-01 | 6 项评测覆盖 100% | 74.5 |
| 03 | Claude Fable 5Anthropic | 2026-06-09 | 6 项评测覆盖 100% | 73.6 |
| 04 | Claude Opus 5.5Anthropic | 2026-09-22 | 6 项评测覆盖 100% | 72.7 |
| 05 | Gemini 4 ArgonGoogle | 2026-09-30 | 2 项评测覆盖 40% | 72.4 |
| 06 | Claude Opus 5Anthropic | 2026-07-24 | 6 项评测覆盖 100% | 71.9 |
| 07 | GPT-6.1 SolOpenAI | 2026-09-29 | 6 项评测覆盖 100% | 71.8 |
| 08 | Gemini 3.8 FlashGoogle | 2026-09-02 | 5 项评测覆盖 80% | 71.1 |
| 09 | Muse Spark 1.3Meta | 2026-09-02 | 4 项评测覆盖 80% | 68.1 |
| 10 | Kimi K3Moonshot AI开源权重 ↗ | 2026-07-16 | 6 项评测覆盖 100% | 68.0 |
| 11 | GPT-5.6 SolOpenAI | 2026-07-09 | 6 项评测覆盖 100% | 65.6 |
| 12 | Claude Opus 4.8Anthropic | 2026-05-28 | 6 项评测覆盖 100% | 65.5 |
| 13 | Gemini 3.7 FlashGoogle | 2026-08-13 | 6 项评测覆盖 100% | 65.3 |
| 14 | GPT-6 SolOpenAI | 2026-09-22 | 6 项评测覆盖 100% | 64.0 |
| 15 | Claude Sonnet 5.5Anthropic | 2026-09-28 | 6 项评测覆盖 100% | 63.5 |
| 16 | GPT-5.5OpenAI | 2026-04-23 | 6 项评测覆盖 100% | 63.3 |
| 17 | Gemini 3.1 Pro PreviewGoogle | 2026-02-19 | 6 项评测覆盖 100% | 63.3 |
| 18 | Grok 4.6xAI | 2026-08-12 | 6 项评测覆盖 100% | 62.5 |
| 19 | GLM-5.3Z.ai开源权重 ↗ | 2026-08-18 | 6 项评测覆盖 100% | 62.4 |
| 20 | Muse Spark 1.2Meta | 2026-08-05 | 6 项评测覆盖 100% | 62.3 |
| 21 | Claude Opus 4.7Anthropic | 2026-04-16 | 4 项评测覆盖 100% | 62.2 |
| 22 | Grok 4.7xAI | 2026-09-21 | 5 项评测覆盖 80% | 62.1 |
| 23 | Gemini 3.6 FlashGoogle | 2026-07-21 | 6 项评测覆盖 100% | 61.0 |
| 24 | Grok 4.5xAI | 2026-07-08 | 6 项评测覆盖 100% | 60.7 |
| 25 | Gemini 3.5 FlashGoogle | 2026-05-19 | 6 项评测覆盖 100% | 60.6 |
| 26 | Step 5 PreviewStepFun | 2026-09-18 | 3 项评测覆盖 40% | 60.0 |
| 27 | GPT-5.6 TerraOpenAI | 2026-07-09 | 6 项评测覆盖 100% | 59.8 |
| 28 | DeepSeek V4.1 FlashDeepSeek开源权重 ↗ | 2026-09-10 | 5 项评测覆盖 100% | 59.8 |
| 29 | Qwen3.7 MaxAlibaba | 2026-05-19 | 5 项评测覆盖 100% | 59.7 |
| 30 | MiMo-V2.6-ProXiaomi开源权重 ↗ | 2026-09-21 | 3 项评测覆盖 40% | 59.6 |
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。分类评分是模型在这个领域的能力，只列这个领域测得够多的模型。 价格为 API 参考报价，原始出处见模型详情。