PlanningBench - Avg-pass: leaderboard

PlanningBench (2026) generates self-contained planning problems from expert-abstracted real scenarios across more than 30 task types, each verified against a checklist of constraints. This board is the average share of checklist items a plan satisfies; percent, higher is better.

Metric: Average checklist pass rate (%): share of an instance's verification checklist items (general, task-specific and stateful constraints) the plan satisfies, averaged over the PlanningBench evaluation instances; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.4xHigh92.35#135
2Gemini 3.1 Pro (Preview)88.36#99
3Seed 2.0 ProHigh84.02#222
4DeepSeek V3.2Thinking80.21#419
5Hy3-preview78.14#322
6Qwen 3.5 PlusThinking77.1#242
7DeepSeek V3.2 Exp76.69#439
8Gemini 2.5 Pro76.41#276
9DeepSeek R174.04#482
10Seed 1.871.69#252
11Hy2.171.18
12Qwen 3 30B A3B57.4#1156
13Qwen 3 32B30.11#972
14Qwen 3 14B25.69#1269
15Qwen 3 8B22.79#1873

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

Interactive version: theaggregate.ai/benchmark?slug=planningbench-avg-pass · How It Works · Data refreshed daily, snapshot 2026-11-21.