PrepBench - Code-to-Workflow Translation: leaderboard
Metric: Translation accuracy (%): given the ground-truth preparation code, the agent must translate it into an executable GUI workflow with the same output, PrepBench natural-language data-preparation tasks, LLM agent at temperature 0.7 with Clarify (budgeted, answered by a DeepSeek-V3.2 user simulator), Profile, Code and Translate actions; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 Codex | 67.7 |
| 2 | Gemini 3 Flash | 52.3 |
| 3 | Kimi K2 (Thinking) | 46.1 |
| 4 | Grok Code Fast 1 | 45.1 |
| 5 | Claude Sonnet 4.5 | 44.8 |
| 6 | Qwen 3 235B A22B | 35 |
| 7 | GLM-4.7 | 34.6 |
| 8 | DeepSeek V3.2 | 27.8 |
| 9 | GPT-4o | 19.9 |
| 10 | Devstral 2 | 10.5 |
Interactive version: theaggregate.ai/benchmark?slug=prepbench-code-to-workflow-translation · How It Works · Data refreshed daily, snapshot 2026-10-07.