Final-Answer Comps — Overall
Math · 2025-02-15
MathArena's current Final-Answer Competitions aggregate is the arithmetic mean of AIME 2026, HMMT February 2026, Apex, and Apex Shortlist scores. It is diagnostic because its four constituent tracks already supply the measured evidence to the capability model.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 94.3 |
| Opus 4.8 | 91.8 |
| Kimi K3 | 87.8 |
| Gemini 3.1 Pro Preview | 86.5 |
| GPT-5.4 | 83.1 |
| Opus 4.6 | 78.5 |
| DeepSeek-V4-Pro | 76.6 |
| DeepSeek-V4-Flash | 76.5 |
| Gemini 3.5 Flash | 76.3 |
| Opus 4.7 | 73.6 |
| Kimi K2.6 | 72.9 |
| GPT-5.2 | 72.0 |
| Gemini 3.6 Flash | 70.8 |
| Step 3.7 Flash | 68.5 |
| Gemini 3 Flash Preview | 67.6 |
| GLM-5.2 | 67.6 |
| GLM-5.1 | 67.1 |
| Gemini 3 Pro Preview | 67.0 |
| Step 3.5 Flash | 66.8 |
| GLM-5 | 65.7 |
| Kimi K2.5 | 62.3 |
| Grok 4.1 Fast | 60.9 |
| Nemotron 3 Super 120B A12B | 60.4 |
| DeepSeek-V3.2 | 57.7 |
| Qwen3.5-27B | 56.7 |
| Qwen3.5 35B A3B | 56.0 |
| Qwen3.5-9B | 48.5 |
| Qwen3 30B A3B Thinking 2507 | 47.8 |
| QED-Nano | 42.7 |
| Qwen3 4B Thinking 2507 | 38.5 |