Smolagents - Overall (Code Agent): leaderboard
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.5 (Preview) | 78.75 |
| 2 | Claude 3.7 Sonnet (20250219) | 77.33 |
| 3 | O1 | 76.67 |
| 4 | O3 Mini | 74.96 |
| 5 | Claude 3.5 Sonnet | 74.75 |
| 6 | DeepSeek R1 | 72.67 |
| 7 | QwQ-32B | 68.5 |
| 8 | GPT-4o | 64.13 |
| 9 | Qwen 2.5 72B Instruct | 64.04 |
| 10 | Llama 3.3 70B Instruct | 60.42 |
| 11 | Qwen 2.5 Coder 32B Instruct | 59.96 |
| 12 | DeepSeek R1 Distill Qwen 32B | 54.42 |
| 13 | Llama 3.1 70B Instruct | 49.88 |
| 14 | Llama 3.2 3B Instruct | 21.04 |
| 15 | Llama 3.1 8B Instruct | 5.33 |
Interactive version: theaggregate.ai/benchmark?slug=smolagents-overall-code-agent · How It Works · Data refreshed daily, snapshot 2026-10-09.