实时基准测试浏览器

在单个交互式视图中查看 Reviuws 记录的每一次基准测试运行 — 按测试、模态和供应商筛选，然后按准确率得分、实测延迟或每百万 token 输入价格重新对排行榜进行排序。

Benchmarks covered

• Reasoning
• Coding
• Instruction Following
• Factuality
• MMLU-Pro
• GPQA Diamond
• Humanity's Last Exam
• AIME 2024
• SWE-bench Verified
• LiveCodeBench
• Terminal-Bench
• HumanEval
• MMMU
• MathVista
• IFEval
• MGSM
• LMArena (Chatbot Arena) Elo
• ARC-AGI-2

Top scoring models

1. Gemini 3 Pro — google
2. Veo 3.1 — google
3. GPT-4o — openai
4. o3 — openai
5. Claude Haiku 4.5 — anthropic
6. DeepSeek-R1 — deepseek
7. Claude Sonnet 4.5 — anthropic
8. Gemini 2.5 Flash — google
9. Claude Opus 4.5 — anthropic
10. Sora 2 — openai
11. Grok 4 — xai
12. Mistral Medium 3 — mistral
13. GPT-5 — openai
14. Llama 4 Maverick — meta
15. Kimi K2 — moonshot ai
16. FLUX.1.1 [pro] — black forest labs
17. Midjourney V7 — midjourney
18. DeepSeek-V3.2 — deepseek
19. Qwen3-Max — alibaba
20. GPT-5.5 — openai
21. SDXL — stability
22. Runway Gen-4 — runway
23. Whisper Large v3 — openai
24. text-embedding-3-small — openai
25. o4-mini — openai