OpenRouter · 研究生水平科学题目(GPQA Diamond)
定期获取并展示公开的基准测试资料(LMArena 用户投票排名、Epoch AI 测试分数、OpenRouter 使用量排名)。每个榜单均标注数据发布日期和获取日期。这些分数并非本站自行评定。
如何解读此排行榜
衡量什么: 这是 OpenRouter 自行运行 GPQA Diamond(生物、化学、物理研究生水平的选择题)测得的正确率。同时公布每题平均成本。
如何解读分数: 正确率(%)越高越好。“每题平均成本”是按 OpenRouter 价格计算的解答一道题的平均费用,可用于比较性能与成本。
注意事项: 运行条件与 Epoch AI 的 GPQA 评测不同,因此得分可能不同。成本按 OpenRouter 价格计算。
本次排名的解读要点
- 最高分为 Gemini 3.8 Flash 的 95.6%。
- 第 1 名与第 5 名的差距为 1.2%p。
- 前 10 个模型中,OpenAI 的模型有 5 个,数量最多。
- 前 10 个模型中,每题平均成本最低的是 Jev Router($0.0159,得分 93.9%)。
研究生水平科学题目(GPQA Diamond) OpenRouter
发布日期:2026-10-04 · 获取日期:2026-10-05
图表仅显示同一模型各推理设置(high、max 等)中排名最高的一项。表格列出所有设置各自的排名。
查看表格(前 60 名)
| 排名 | 模型 | 开发方 | 分数 | 每题平均成本 |
|---|---|---|---|---|
| 1 | #1Gemini 3.8 Flash | 95.6% | $0.0682 | |
| 2 | #2OGPT-6 Astra Pro | OpenAI | 95.5% | $0.3326 |
| 3 | #3SFugu Ultra | Sakana | 94.6% | $1.1487 |
| 4 | #4Gemini 3.1 Pro Preview | 94.5% | $0.202 | |
| 5 | #5OGPT-6 Astra | OpenAI | 94.4% | $0.1133 |
| 6 | #6OGPT-6.1 Sol | OpenAI | 94.4% | $0.0214 |
| 7 | #7TJev Router | Typesafe | 93.9% | $0.0159 |
| 8 | #8OGPT-5.6 Sol Pro | OpenAI | 93.9% | $0.2979 |
| 9 | #9Gemini 3.7 Flash | 93.9% | $0.0293 | |
| 10 | #10OGPT-5.5 | OpenAI | 93.7% | $0.3093 |
| 11 | #11SFugu Ultra V2 | Sakana | 93.3% | $0.5026 |
| 12 | #12xGrok 4.6 | xAI | 92.8% | $0.1233 |
| 13 | #13Gemini 3.6 Flash | 92.6% | $0.0596 | |
| 14 | #14Gemini 3.5 Flash | 92.6% | $0.1413 | |
| 15 | #15UPareto | Unbiased | 92.4% | $0.0338 |
| 16 | #16OGPT-5.6 Sol | OpenAI | 92.1% | $0.0604 |
| 17 | #17Claude Sonnet 5.5 | Anthropic | 92.1% | $0.0253 |
| 18 | #18OGPT-6 Sol | OpenAI | 91.9% | $0.0253 |
| 19 | #19Kimi K3 | Moonshot AI | 91.9% | $0.1323 |
| 20 | #20Switchyard | NVIDIA | 91.4% | $0.0179 |
| 21 | #21ZGLM 5.3 Flash | Z.ai | 90.9% | $0.0054 |
| 22 | #22Qwen3.7 Max | Alibaba | 90.9% | $0.2222 |
| 23 | #23MiniMax M3 | MiniMax | 90.3% | $0.0348 |
| 24 | #24OGPT-5.4 | OpenAI | 90.3% | $0.1467 |
| 25 | #25Claude Opus 5.5 | Anthropic | 90.2% | $0.0542 |
| 26 | #26OGPT-5.6 Luna Pro | OpenAI | 90% | $0.0412 |
| 27 | #27Claude Fable 5.1 | Anthropic | 89.9% | $0.1305 |
| 28 | #28DeepSeek V4 Pro 0813 | DeepSeek | 89.8% | $0.1143 |
| 29 | #29Claude Opus 4.8 | Anthropic | 89.7% | $0.2016 |
| 30 | #30Claude Opus 4.7 | Anthropic | 89.4% | $0.2239 |
| 31 | #31THy4 preview | Tencent | 89.1% | $0.1253 |
| 32 | #32ANova Micro 1.0 | Amazon | 89% | $0.2165 |
| 33 | #33DeepSeek V4.1 Flash | DeepSeek | 88.9% | $0.0244 |
| 34 | #34OGPT-5.6 Terra | OpenAI | 88.8% | $0.0344 |
| 35 | #35MiMo-V2.6-Pro | Xiaomi | 88.6% | $0.0319 |
| 36 | #36SFugu Max | Sakana | 88.6% | $0.0756 |
| 37 | #37Qwen3.8 Flash | Alibaba | 88.6% | $0.0215 |
| 38 | #38Gemini 3 Flash Preview | 88.4% | $0.1318 | |
| 39 | #39OGPT-6 Luna Pro | OpenAI | 88.4% | $0.0083 |
| 40 | #40DeepSeek V4 Flash Vision Exp | DeepSeek | 88% | $0.0228 |
| 41 | #41OAuto Router | Openrouter | 88% | $0.0892 |
| 42 | #42OGPT-5.2 | OpenAI | 87.9% | $0.1333 |
| 43 | #43Qwen3.7 Plus | Alibaba | 87.9% | $0.044 |
| 44 | #44OGPT-5.6 Luna | OpenAI | 87.9% | $0.0079 |
| 45 | #45Claude Opus 5 | Anthropic | 87.5% | $0.1306 |
| 46 | #46OGPT-6 Luna | OpenAI | 87.4% | $0.003 |
| 47 | #47Kimi K2 Thinking | Moonshot AI | 87.3% | $0.1062 |
| 48 | #48Qwen3.8 2.4T A95B | Alibaba | 86.9% | $0.0824 |
| 49 | #49TInkling Small | Thinkingmachines | 86.6% | $0.0253 |
| 50 | #50OGPT-5.1 | OpenAI | 86.4% | $0.2213 |
| 51 | #51DeepSeek V4 Flash 0731 | DeepSeek | 86.3% | $0.008 |
| 52 | #52MiniMax M2.1 | MiniMax | 86.2% | $0.0403 |
| 53 | #53Claude Opus 4.5 | Anthropic | 86.2% | $0.8454 |
| 54 | #54DeepSeek V4 Pro 0423 | DeepSeek | 86.2% | $0.0544 |
| 55 | #55Claude Opus 4.6 | Anthropic | 86% | $0.7469 |
| 56 | #56DeepSeek V4 Flash 0423 | DeepSeek | 85.8% | $0.0062 |
| 57 | #57ZGLM 5.3 Flash | Z.ai | 85.7% | $0.0123 |
| 58 | #58OGPT-5 | OpenAI | 85.7% | $0.2099 |
| 59 | #59Claude Fable 5 | Anthropic | 85.5% | $0.1754 |
| 60 | #60ZGLM 5.3 | Z.ai | 85.4% | $0.0925 |
Source: OpenRouter evals (openrouter.ai) via OpenRouter (openrouter.ai/rankings). Licensed under CC BY 4.0.
数据:LMArena leaderboard dataset(CC BY 4.0)。修改:各榜单仅显示前 50 名,分数四舍五入。 https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset公司标志为各公司的商标,仅用于帮助区分(图标:Simple Icons)。数据:Epoch AI — Capabilities & benchmarking(CC BY 4.0)。 https://epoch.ai/benchmarksSource: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings). Source: OpenRouter evals (openrouter.ai) via OpenRouter (openrouter.ai/rankings). Source: Design Arena (www.designarena.ai) via OpenRouter (openrouter.ai/rankings). Licensed under CC BY 4.0. Source: OpenRouter (openrouter.ai/rankings), as of 2026-10-05. Licensed under CC BY 4.0. https://openrouter.ai/rankings