Plan-RewardBench - Single-Turn Planning (Hard): leaderboard

Plan-RewardBench (2026) tests reward models and LLM judges on preferring the better of two tool-using agent trajectories, with preference labels from a judge panel audited by humans. This board is pairwise accuracy in percent on the hard single-turn planning pairs; higher is better.

Metric: Pairwise accuracy (%) on the 158 hard single-turn planning pairs of the Plan-RewardBench chosen versus rejected agent trajectories (shared tool registry, multi-turn context, executed tool feedback), preference labels from a Gemini 3 Pro and GPT-5.1 judge panel with human audit; discriminative reward models score each trajectory, generative reward models and LLM judges choose between the pair; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1DeepSeek V3.2 Exp74.84#440
2Qwen 3 30B A3B 2507 Instruct74.68#1106
3Qwen Plus74.68#828
4INF-ORM-Llama3.1-70B74.05
5InternLM2-7B-Reward74.03
6Qwen 3 4B 2507 Instruct73.73#2240
7DeepSeek R173.58#484
8FsfairX-LLaMA3-RM-v0.172.78
9Skywork-Reward-V2-Llama-3.1-8B72.15
10Qwen 3 235B A22B 2507 Instruct71.36#625
11RRM-32B70.8#1127
12Skywork-Reward-V2-Qwen3-8B70.57
13Qwen Max68.99#1094
14RM-R1-DS-Distill-Qwen-32B68.57
15M-Prometheus-14B68.4#1263

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=plan-rewardbench-single-turn-planning-hard · How It Works · Data refreshed daily, snapshot 2026-11-22.