PrepBench - Prep-Code Generation: leaderboard
Metric: Prep-code accuracy (%) from the disambiguated request with a data sample and profiling, output table compared with the reference, 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 | 85.3 |
| 2 | Gemini 3 Flash | 74.8 |
| 3 | Claude Sonnet 4.5 | 70.3 |
| 4 | Kimi K2 (Thinking) | 69.3 |
| 5 | Qwen 3 235B A22B | 67.7 |
| 6 | GLM-4.7 | 65.7 |
| 7 | DeepSeek V3.2 | 65.4 |
| 8 | Grok Code Fast 1 | 57.5 |
| 9 | Devstral 2 | 44.1 |
| 10 | GPT-4o | 39.5 |
Interactive version: theaggregate.ai/benchmark?slug=prepbench-prep-code-generation · How It Works · Data refreshed daily, snapshot 2026-10-07.