CodeEditorBench Plus - Code Translate (Three-shot): leaderboard
Code Translate task of CodeEditorBench (Guo et al., 2024) on its Plus split, the problems kept after timestamp-based filtering against training-data contamination: the model translates a program between C++, Java and Python, and passes when the translation passes every test case in the target language. 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) | 51.7 | #1214 |
| 2 | Gemini 1.0 Pro | 39.2 | #4110 |
| 3 | GPT-3.5 Turbo (1106) | 36.4 | #3245 |
| 4 | CodeLlama-13B-Instruct-hf | 32.7 | #9291 |
| 5 | Magicoder-S-DS-6.7B | 32.1 | #7023 |
| 6 | CodeLlama-34B-hf | 30.7 | #9332 |
| 7 | CodeLlama-34B-Instruct-hf | 30.3 | #7428 |
| 8 | GLM-4 | 29.9 | #1673 |
| 9 | Phind-CodeLlama-34B-v2 | 27.5 | #9928 |
| 10 | CodeLlama-7B-Instruct-hf | 27.1 | #9444 |
| 11 | WizardCoder-15B-V1.0 | 27.1 | #9717 |
| 12 | Magicoder-S-CL-7B | 24.5 | #9042 |
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-translate-three-shot · How It Works · Data refreshed daily, snapshot 2026-11-12.