RMGAP - Pointwise LLM Judge: leaderboard

Metric: Pairwise accuracy (%; chosen response scored above a rejected one, prompts asking for one of four styles, mean of chat, writing, reasoning and safety; the model rates each response 1-10, 5 when no score can be parsed, ties count as errors). Source: arxiv.org. Saturation forecast: Around December 2026. 5 models tracked.

Top models

#ModelScore
1Qwen 3 235B A22B 250758.08
2DeepSeek V3.252.51
3Qwen 3 8B44.95
4Qwen 2.5 7B Instruct43.77
5Llama 3.1 8B Instruct30

Interactive version: theaggregate.ai/benchmark?slug=rmgap-pointwise-llm-judge · How It Works · Data refreshed daily, snapshot 2026-09-26.