CodeEditorBench Plus - Code Requirement Switch (Three-shot): leaderboard
Code Requirement Switch task of CodeEditorBench (Guo et al., 2024) on its Plus split, the problems kept after timestamp-based filtering against training-data contamination: the model adapts a solution of a similar problem to a new problem statement, and passes when the adapted program passes every test case. Scored as pass@1 (%) with greedy decoding and three in-context examples; higher is better.
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 | Overall rank |
|---|---|---|---|
| 1 | GPT-4 (0613) | 30.3 | #1214 |
| 2 | GPT-3.5 Turbo (1106) | 20.1 | #3245 |
| 3 | Magicoder-S-DS-6.7B | 19.2 | #7023 |
| 4 | Gemini 1.0 Pro | 13.9 | #4110 |
| 5 | CodeLlama-34B-hf | 11.3 | #9332 |
| 6 | GLM-4 | 10 | #1673 |
| 7 | WizardCoder-15B-V1.0 | 9.9 | #9717 |
| 8 | Phind-CodeLlama-34B-v2 | 9.2 | #9928 |
| 9 | Magicoder-S-CL-7B | 7.5 | #9042 |
| 10 | CodeLlama-34B-Instruct-hf | 3.2 | #7428 |
| 11 | CodeLlama-7B-Instruct-hf | 2.8 | #9444 |
| 12 | CodeLlama-13B-Instruct-hf | 2.8 | #9291 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=codeeditorbench-plus-code-requirement-switch-three-shot · How It Works · Data refreshed daily, snapshot 2026-11-12.