RubricBench - Code: leaderboard
Metric: Preference accuracy (%) on the code pairs on RubricBench's 1,147 preference pairs (instruction following, STEM, code, safety and chat pairs re-curated from HelpSteer3, PPE and RewardBench2 and filtered so that surface cues such as length, formatting or tone favour the rejected response); the reward model or LLM judge picks the preferred response directly, with no rubric; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-3.1-Tulu-3-8B-RM | 53.5 | #790 |
| 2 | GPT-4o Mini | 51.9 | #588 |
| 3 | ArmoRM-Llama3-8B-v0.1 | 50.2 | #1375 |
| 4 | DeepSeek V3.2 (Non-reasoning) | 46.9 | #198 (DeepSeek V3.2) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=rubricbench-code · How It Works · Data refreshed daily, snapshot 2026-10-11.