NewtonBench - Code Assistance: leaderboard
Metric: Symbolic Accuracy (%; 324 tasks, mean of four runs). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 73.3 |
| 2 | Gemini 2.5 Pro | 63.7 |
| 3 | GPT-5 Mini | 48.1 |
| 4 | O4 Mini | 45.1 |
| 5 | Gemini 2.5 Flash | 40.9 |
| 6 | DeepSeek R1 | 37.4 |
| 7 | Qwen 3 235B A22B | 33.7 |
| 8 | QwQ-32B | 31.4 |
| 9 | GPT-4.1 | 18.9 |
| 10 | GPT-4.1 Mini | 14.7 |
| 11 | DeepSeek V3 | 13.3 |
Interactive version: theaggregate.ai/benchmark?slug=newtonbench-code-assistance · How It Works · Data refreshed daily, snapshot 2026-09-24.