最佳AI翻译模型
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. Claude Mythos Preview (Anthropic, 分数: 98.9), 2. Gemini 3.1 Pro (Google, 分数: 79.3), 3. Claude Fable 5.1 (Anthropic, 分数: 77.2).
| # | 模型 | 提供商 | 分数 | 速度 | 价格（输出） |
|---|---|---|---|---|---|
| 1 | Claude Mythos Preview | Anthropic | 98.9 | 80 t/s | $15/M |
| 2 | Gemini 3.1 Pro | 79.3 | 136 t/s | $12/M | |
| 3 | Claude Fable 5.1 | Anthropic | 77.2 | 85 t/s | $15/M |
| 4 | Claude Mythos 5 | Anthropic | 75.0 | 50 t/s | $50/M |
| 5 | Claude Sonnet 4.6 (max) | Anthropic | 67.0 | 50 t/s | $15/M |
| 6 | Qwen3.8 Flash Next | Alibaba | 63.9 | 168 t/s | $0.4/M |
| 7 | Claude Fable 5.1 (XHigh) | Anthropic | 61.5 | 78 t/s | $15/M |
| 8 | GPT-6 Astra (xhigh) | OpenAI | 61.5 | 76 t/s | $50/M |
| 9 | Claude Fable 5.1 (High) | Anthropic | 61.5 | 92 t/s | $15/M |
| 10 | GPT-6 Astra (high) | OpenAI | 61.5 | 90 t/s | $50/M |
| 11 | Claude Fable 5.1 (Medium) | Anthropic | 61.5 | 104 t/s | $15/M |
| 12 | GPT-6 Astra (medium) | OpenAI | 61.5 | 98 t/s | $50/M |
| 13 | Claude Fable 5.1 (Low) | Anthropic | 61.5 | 118 t/s | $15/M |
| 14 | GPT-6 Astra (low) | OpenAI | 61.5 | 110 t/s | $50/M |
| 15 | Claude Opus 5.5 | Anthropic | 60.9 | 68 t/s | $25/M |
Claude Mythos Preview
Gemini 3.1 Pro
Qwen3.8 Flash Next
GPT-6 Astra (xhigh)