Apex Shortlist
Math · 2025-10-25
This MathArena track evaluates models on the Apex Shortlist problem set. Scores report the percentage of problems answered correctly.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 98.4 |
| Gemini 3.1 Pro Preview | 92.0 |
| Kimi K3 | 92.0 |
| Opus 4.8 | 90.4 |
| DeepSeek-V4-Flash | 89.4 |
| DeepSeek-V4-Pro | 87.8 |
| Opus 4.6 | 86.7 |
| Gemini 3.5 Flash | 82.5 |
| GPT-5.4 | 81.4 |
| GPT-5.2 | 79.3 |
| Kimi K2.6 | 77.1 |
| Step 3.7 Flash | 76.6 |
| GLM-5.1 | 71.8 |
| Gemini 3.6 Flash | 71.3 |
| Step 3.5 Flash | 70.7 |
| DeepSeek-V3.2-Speciale | 69.7 |
| Gemini 3 Flash Preview | 68.6 |
| GLM-5 | 68.6 |
| GLM-5.2 | 68.1 |
| Gemini 3 Pro Preview | 66.5 |
| Opus 4.7 | 63.8 |
| Qwen3.5 397B A17B | 60.1 |
| Grok 4.1 Fast | 57.9 |
| Grok 4 | 57.5 |
| Kimi K2.5 | 57.5 |
| Nemotron 3 Super 120B A12B | 57.5 |
| GPT-5.1 | 55.9 |
| Grok 4 Fast | 55.3 |
| Qwen3.5-27B | 52.1 |
| DeepSeek-V3.2 | 50.5 |
| Kimi K2 Thinking | 47.3 |
| gpt-oss-120b | 45.2 |
| Qwen3.5 35B A3B | 44.7 |
| GPT-5 Mini | 39.9 |
| Qwen3.5-4B | 32.5 |
| Qwen3.5-9B | 29.8 |
| Qwen3 30B A3B Thinking 2507 | 23.4 |
| QED-Nano | 22.3 |
| Qwen3 4B Thinking 2507 | 16.5 |
| Qwen3.5-2B | 9.0 |