Agent Planning Benchmark - Tool-Extraneous Planning: leaderboard
The Agent Planning Benchmark (APB, 2026) isolates the planning step of multimodal tool-using agents: models write plans and tool calls for 4,209 cases drawn from agent datasets and real platform traffic, and Gemini 3 Pro judges each plan against a reference trajectory. This board adds 2 to 10 plausible but irrelevant tools to the tool set in 750 holistic and 750 step-wise cases and reports the percentage of correct plans; higher is better.
Metric: Correctness rate (%) when 2, 4, 6, 8 or 10 semantically similar but irrelevant tools are added to the tool set, over 1,500 instances (750 holistic and 750 step-wise, 150 per distractor level); binary plan correctness judged by Gemini 3 Pro against a reference trajectory (92.2% agreement with human annotation); multimodal tool-use cases aggregated from OpenCUA, GTA, GAIA, ToolBench, FrameThinker and real agent-platform traffic; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 76.4 | #250 |
| 2 | Gemini 3 Pro (Preview) (High) | 76.4 | #113 (Gemini 3 Pro (Preview)) |
| 3 | GPT-5 | 67 | #167 |
| 4 | Gemini 2.5 Pro | 66.8 | #266 |
| 5 | Gemini 2.5 Flash (Non-reasoning) | 59.8 | #453 (Gemini 2.5 Flash) |
| 6 | Qwen 3 VL 235B A22B Instruct | 52.6 | #522 |
| 7 | Qwen 3 VL 32B Instruct | 51.9 | #585 |
| 8 | GPT-4o | 45.5 | #741 |
| 9 | Qwen 3 VL 30B A3B Instruct | 40.2 | #871 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=agent-planning-benchmark-tool-extraneous-planning · How It Works · Data refreshed daily, snapshot 2026-11-13.