最佳文档分析AI
排名结合多个基准分数，并按与此用例的相关性加权，帮助你比较适合不同任务的模型。更新于2026年8月。
快速答案
1. Qwen3.7 Max (Alibaba, 分数: 82.6), 2. GPT-6 Astra (xhigh) (OpenAI, 分数: 81.5), 3. GPT-6 Astra (high) (OpenAI, 分数: 81.5).
| # | 模型 | 提供商 | 分数 | 速度 | 价格（输出） |
|---|---|---|---|---|---|
| 1 | Qwen3.7 Max | Alibaba | 82.6 | 203 t/s | $7.5/M |
| 2 | GPT-6 Astra (xhigh) | OpenAI | 81.5 | 76 t/s | $50/M |
| 3 | GPT-6 Astra (high) | OpenAI | 81.5 | 90 t/s | $50/M |
| 4 | GPT-6 Astra (medium) | OpenAI | 81.5 | 98 t/s | $50/M |
| 5 | GPT-6 Astra (low) | OpenAI | 81.5 | 110 t/s | $50/M |
| 6 | Claude Fable 5 | Anthropic | 78.8 | 71 t/s | $50/M |
| 7 | GPT-6 Astra | OpenAI | 78.7 | 82 t/s | $50/M |
| 8 | Claude Opus 5 | Anthropic | 77.2 | 60 t/s | $25/M |
| 9 | Claude Mythos 5 | Anthropic | 76.7 | 50 t/s | $50/M |
| 10 | Claude Opus 5.5 | Anthropic | 75.7 | 68 t/s | $25/M |
| 11 | Qwen3.8 Flash Next | Alibaba | 75.4 | 168 t/s | $0.4/M |
| 12 | Claude Fable 5.1 | Anthropic | 74.8 | 85 t/s | $15/M |
| 13 | Claude Opus 4.8 | Anthropic | 74.8 | 50 t/s | $15/M |
| 14 | GLM-5.3 Flash | Z.ai | 74.7 | 182 t/s | $0.3/M |
| 15 | DeepSeek V4.1 Flash | DeepSeek | 74.4 | 214 t/s | $1.2/M |
Qwen3.7 Max
GPT-6 Astra (xhigh)
Claude Fable 5
GLM-5.3 Flash
DeepSeek V4.1 Flash