Plan-RewardBench - Single-Turn Planning (Easy): leaderboard
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.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen Plus | 84.55 |
| 2 | Qwen 3 235B A22B 2507 Instruct | 84.38 |
| 3 | GPT-5 | 83.85 |
| 4 | Qwen 3 30B A3B 2507 Instruct | 82.47 |
| 5 | Gemini 3 Flash (Preview) | 81.08 |
| 6 | Qwen Max | 80.73 |
| 7 | InternLM2-7B-Reward | 79.86 |
| 8 | INF-ORM-Llama3.1-70B | 79.86 |
| 9 | DeepSeek V3.2 Exp | 79.51 |
| 10 | DeepSeek R1 | 76.74 |
| 11 | Qwen 3 4B 2507 Instruct | 76.56 |
| 12 | Kimi K2 (Thinking) | 73.02 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 72.4 |
| 14 | Skywork-Reward-V2-Llama-3.1-8B | 69.79 |
| 15 | FsfairX-LLaMA3-RM-v0.1 | 68.06 |
Interactive version: theaggregate.ai/benchmark?slug=plan-rewardbench-single-turn-planning-easy · How It Works · Data refreshed daily, snapshot 2026-10-07.