RedCode - Exec Python (CodeAct): leaderboard
Metric: Attack success rate (%). Source: redcode-agent.github.io. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | codellama-7b-instruct-hf | 61.83 |
| 2 | CodeActAgent-Mistral-7b-v0.1 | 62.6 |
| 3 | CodeActAgent-Llama-2-7b | 69.95 |
| 4 | codellama-13b-instruct-hf | 71.87 |
Interactive version: theaggregate.ai/benchmark?slug=redcode-exec-python-codeact · How It Works · Data refreshed daily, snapshot 2026-10-09.