CodeEditorBench Plus - Code Debug (Three-shot): leaderboard
Metric: pass@1 (%, greedy decoding, three-shot, CodeEditorBench_Plus). Source: codeeditorbench.github.io. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 (0613) | 34.5 |
| 2 | GPT-3.5 Turbo (1106) | 27 |
| 3 | Phind-CodeLlama-34B-v2 | 23.9 |
| 4 | GLM-4 | 23.3 |
| 5 | Gemini 1.0 Pro | 22.9 |
| 6 | CodeLlama-7B-Instruct-hf | 16.7 |
| 7 | CodeLlama-13B-Instruct-hf | 16 |
| 8 | Magicoder-S-CL-7B | 15.7 |
| 9 | CodeLlama-34B-Instruct-hf | 14.3 |
| 10 | CodeLlama-34B-hf | 13.3 |
| 11 | WizardCoder-15B-V1.0 | 11.4 |
Interactive version: theaggregate.ai/benchmark?slug=codeeditorbench-plus-code-debug-three-shot · How It Works · Data refreshed daily, snapshot 2026-09-26.