Benchmarks · math
MathArena Apex Shortlist
MathArena's Apex shortlist: the hardest final-answer problems from 2025 competitions that models still struggle with. Accuracy over repeated attempts.
33 models · 33 results · data as of 2026-10-29 · matharena.ai ↗
Leaderboard
GPT-5.5 xhigh
All results
33 rows| # | Model | Effort | MathArena Apex Shortlist | Subscription $ / task | API $ / task |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | xhigh | 98.4% ±1.7 | $0.043 | $0.77 |
| 2 | Gemini 3.1 Pro | high | 92.0% ±3.9 | $0.37 | $0.37 |
| 3 | Kimi K3 | max | 92.0% ±3.9 | $0.051 | $0.66 |
| 4 | Opus 4.8 | max | 90.4% ±4.2 | $0.054 | $3.14 |
| 5 | DeepSeek V4 Flash | max | 89.4% ±4.4 | $0.002 | $0.040 |
| 6 | DeepSeek V4 Pro | max | 87.8% ±4.7 | $0.057 | $0.34 |
| 7 | Opus 4.6 | high | 86.7% ±4.9 | $1.80 | $1.80 |
| 8 | Gemini 3.5 Flash | xhigh | 82.5% ±5.4 | $0.26 | $0.26 |
| 9 | GPT-5.4 | xhigh | 81.4% ±5.6 | $0.53 | $0.53 |
| 10 | GPT-5.2 | high | 79.3% ±5.8 | $0.76 | $0.76 |
| 11 | Kimi K2.6 | max | 77.1% ±6.0 | $0.27 | $0.27 |
| 12 | GLM 5.1 | max | 71.8% ±6.4 | $0.24 | $0.24 |
| 13 | Gemini 3.6 Flash | xhigh | 71.3% ±6.5 | $0.19 | $0.19 |
| 14 | DeepSeek V3.2 | high | 69.7% ±6.6 | $0.028 | $0.028 |
| 15 | Gemini 3 Flash | high | 68.6% ±6.6 | $0.090 | $0.090 |
| 16 | GLM 5 | high | 68.6% ±6.6 | $0.22 | $0.22 |
| 17 | GLM 5.2 | max | 68.1% ±6.7 | $0.014 | $0.16 |
| 18 | Gemini 3 Pro | high | 66.5% ±6.7 | $0.27 | $0.27 |
| 19 | Opus 4.7 | xhigh | 63.8% ±6.9 | $1.83 | $1.83 |
| 20 | Qwen 3.5 397B-A17B | max | 60.1% ±7.0 | $0.14 | $0.14 |
| 21 | Grok 4.1 Fast | high | 57.9% ±2.5 | $0.014 | $0.014 |
| 22 | Grok 4 | high | 57.5% ±7.1 | $0.52 | $0.52 |
| 23 | Kimi K2.5 | max | 57.5% ±7.1 | $0.16 | $0.16 |
| 24 | GPT-5.1 | high | 55.9% ±7.1 | $0.57 | $0.57 |
| 25 | Grok 4 Fast | none | 55.3% ±7.1 | $0.012 | $0.012 |
| 26 | Qwen 3.5 27B | high | 52.1% ±7.2 | $0.15 | $0.15 |
| 27 | Kimi K2 | medium | 47.3% ±7.1 | $0.14 | $0.14 |
| 28 | gpt-oss-120b | high | 45.2% ±7.1 | $0.026 | $0.026 |
| 29 | Qwen 3.5 35B-A3B | low | 44.7% ±7.1 | $0.085 | $0.085 |
| 30 | GPT-5 Mini | high | 39.9% ±7.0 | $0.068 | $0.068 |