Plan-RewardBench - Single-Turn Planning (Easy): 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 easy single-turn planning pairs; higher is better.

Metric: Pairwise accuracy (%) on the 144 easy 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
1Qwen Plus84.55#820
2Qwen 3 235B A22B 2507 Instruct84.38#616
3GPT-583.85#178
4Qwen 3 30B A3B 2507 Instruct82.47#1100
5Gemini 3 Flash (Preview)81.08#152
6Qwen Max80.73#1092
7InternLM2-7B-Reward79.86
8INF-ORM-Llama3.1-70B79.86
9DeepSeek V3.2 Exp79.51#439
10DeepSeek R176.74#482
11Qwen 3 4B 2507 Instruct76.56#2221
12RRM-32B75.22#1115
13Kimi K2 (Thinking)73.02
14Qwen 3 235B A22B 2507 (Thinking)72.4#466
15M-Prometheus-14B72.34#1255

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=plan-rewardbench-single-turn-planning-easy · How It Works · Data refreshed daily, snapshot 2026-11-21.