Benchmarks · math
MathArena Apex
MathArena Apex: 12 final-answer competition problems that earlier frontier models (Grok 4, GPT-5, Gemini 2.5 Pro, GLM 4.5) failed, so it is biased against them. Accuracy over repeated attempts.
41 models · 41 results · data as of 2026-11-06 · matharena.ai ↗
Leaderboard
Opus 4.8 max
Heavy line: best score at each price. Thin lines join one model's effort levels. Cost uses each model's median subscription plan, or API price with no plan data. 28 more models under All models in settings.
All results
41 rows| # | Model | Effort | MathArena Apex | Subscription $ / task | API $ / task |
|---|---|---|---|---|---|
| 1 | Opus 4.8 | max | 81.3% ±7.8 | $0.080 | $4.59 |
| 2 | GPT-5.5 | xhigh | 80.2% ±8.0 | $0.039 | $1.42 |
| 3 | GPT-5.4 Pro | xhigh | 69.8% ±9.2 | $8.65 | $8.65 |
| 4 | Kimi K3 | max | 65.6% ±9.5 | $0.10 | $1.11 |
| 5 | Gemini 3.1 Pro | high | 60.9% ±6.9 | $0.41 | $0.41 |
| 6 | GPT-5.4 | xhigh | 54.2% ±7.1 | $1.03 | $1.03 |
| 7 | Opus 4.7 | xhigh | 40.6% ±9.8 | $2.10 | $2.10 |
| 8 | Opus 4.6 | high | 34.5% ±6.8 | $2.36 | $2.36 |
| 9 | Gemini 3.5 Flash | xhigh | 32.3% ±9.3 | $0.30 | $0.30 |
| 10 | DeepSeek V4 Pro | max | 28.1% ±9.0 | $0.070 | $0.42 |
| 11 | DeepSeek V4 Flash | max | 27.1% ±9.3 | $0.003 | $0.050 |
| 12 | Gemini 3.6 Flash | xhigh | 26.0% ±8.8 | $0.006 | $0.19 |
| 13 | Kimi K2.6 | max | 24.0% ±8.5 | $0.16 | $0.32 |
| 14 | Gemini 3 Pro | high | 23.4% ±6.0 | $0.28 | $0.28 |
| 15 | GLM 5.2 | max | 19.8% ±8.0 | $0.005 | $0.18 |
| 16 | Gemini 3 Flash | high | 15.6% ±5.1 | $0.10 | $0.10 |
| 17 | GPT-5.2 | high | 13.5% ±4.8 | $1.00 | $1.00 |
| 18 | GLM 5.1 | max | 11.5% ±4.5 | $0.070 | $0.27 |
| 19 | GLM 5 | high | 10.9% ±4.4 | $0.13 | $0.25 |
| 20 | DeepSeek V3.2 | high | 9.4% ±4.1 | $0.031 | $0.031 |
| 21 | Kimi K2.5 | max | 8.8% ±4.0 | $0.090 | $0.18 |
| 22 | Grok 4 Fast | none | 5.2% ±3.1 | $0.011 | $0.011 |
| 23 | Grok 4.1 Fast | high | 5.2% ±3.1 | $0.013 | $0.013 |
| 24 | Qwen 3 235B | 5.2% ±3.1 | $0.052 | $0.052 | |
| 25 | Qwen 3.5 35B-A3B | low | 4.2% ±2.8 | $0.085 | $0.085 |
| 26 | GPT-5 | high | 2.1% ±3.1 | $3.83 | $3.83 |
| 27 | Grok 4 | high | 2.1% ±2.0 | $0.52 | $0.52 |
| 28 | Qwen 3 4B | none | 2.1% ±2.0 | $0.008 | $0.008 |
| 29 | Qwen 3.5 27B | high | 2.1% ±2.0 | $0.10 | $0.10 |
| 30 | Sonnet 4.5 | high | 1.6% ±1.7 | $0.38 | $0.38 |