AI BENCHY Category
Puzzle Solving Ranking
See which AI models perform best on Puzzle Solving, which ones stay reliable, and where the biggest gaps appear. Sort by: Tests Correct ↓.
| Rank | Model | Company | Puzzle Solving Score | Score | Tests Correct | Response Time (avg) |
|---|---|---|---|---|---|---|
| #37 | Claude Opus 4.6 medium | Anthropic | 7.7 | 7.6 | 2/3 | 4.60s |
| #39 | Seed-2.0-Mini medium | Bytedance Seed | 8.2 | 7.5 | 2/3 | 25.9s |
| #40 | GPT-5.2 medium | OpenAI | 7.7 | 7.5 | 2/3 | 5.47s |
| #41 | MiMo-V2-Flash medium | Xiaomi | 7.7 | 7.5 | 2/3 | 3.77s |
| #42 | Claude Sonnet 4.6 none | Anthropic | 7.7 | 7.4 | 2/3 | 2.92s |
| #49 | Qwen3.5 Plus 2026-02-15 none | Qwen | 7.7 | 6.8 | 2/3 | 2.82s |
| #53 | GLM 5 none | Z.ai | 7.7 | 6.6 | 2/3 | 2.05s |
| #55 | MiMo-V2-Omni none | Xiaomi | 8.0 | 6.5 | 2/3 | 2.71s |
| #64 | DeepSeek V3.2 none | DeepSeek | 8.5 | 6.1 | 2/3 | 7.37s |
| #18 | GLM 5 Turbo medium | Z.ai | 7.3 | 8.1 | 1/3 | 5.44s |
| #23 | MiMo-V2-Pro medium | Xiaomi | 7.0 | 8.1 | 1/3 | 4.71s |
| #30 | Step 3.5 Flash medium | Stepfun | 5.3 | 7.9 | 1/3 | 7.72s |
| #32 | Qwen3.5-Flash medium | Qwen | 6.4 | 7.8 | 1/3 | 56.7s |
| #35 | MiMo-V2-Omni medium | Xiaomi | 6.5 | 7.7 | 1/3 | 3.88s |
| #43 | Qwen3.5-35B-A3B medium | Qwen | 6.4 | 7.4 | 1/3 | 31.6s |