Apex (MathArena)
Math · 2025-08-18
This MathArena track evaluates models on the Apex (MathArena) problem set. Scores report the percentage of problems answered correctly. The leaderboard reports run-averaged accuracy (decimal percentages that land off the 12-item lattice), so the reported number is a mean of per-item pass rates rather than a single-run proportion, and a binomial noise model is not licensed.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.8 | 81.3 |
| GPT-5.5 | 80.2 |
| GPT-5.4 Pro | 69.8 |
| Kimi K3 | 65.6 |
| Gemini 3.1 Pro Preview | 60.9 |
| GPT-5.4 | 54.2 |
| Opus 4.7 | 40.6 |
| Opus 4.6 | 34.5 |
| Gemini 3.5 Flash | 32.3 |
| DeepSeek-V4-Pro | 28.1 |
| DeepSeek-V4-Flash | 27.1 |
| Gemini 3.6 Flash | 26.0 |
| Kimi K2.6 | 24.0 |
| Gemini 3 Pro Preview | 23.4 |
| GLM-5.2 | 19.8 |
| Gemini 3 Flash Preview | 15.6 |
| Step 3.7 Flash | 14.6 |
| GPT-5.2 | 13.5 |
| Step 3.5 Flash | 13.5 |
| GLM-5.1 | 11.5 |
| GLM-5 | 10.9 |
| DeepSeek-V3.2-Speciale | 9.4 |
| Kimi K2.5 | 8.8 |
| Nemotron 3 Super 120B A12B | 7.8 |
| Grok 4.1 Fast | 5.2 |
| Grok 4 Fast | 5.2 |
| Qwen3 235B A22B Thinking 2507 | 5.2 |
| Qwen3.5 35B A3B | 4.2 |
| DeepSeek-V3.2 | 2.1 |
| GPT-5 | 2.1 |
| Grok 4 | 2.1 |
| Qwen3.5-27B | 2.1 |
| Qwen3 4B Thinking 2507 | 2.1 |
| Sonnet 4.5 | 1.6 |
| QED-Nano | 1.6 |
| DeepSeek-R1-0528 | 1.0 |
| GLM-4.5 | 1.0 |
| GPT-5 Mini | 1.0 |
| gpt-oss-120b | 1.0 |
| DeepSeek-V3.1 | 0.5 |