MultiUAV-Plat (ReAct): leaderboard
MultiUAV-Plat (2026) is a benchmark of 1,500 natural-language multi-UAV missions (area search, area assignment and patrol, target assignment) solved through the platform RESTful APIs and checked by hidden validation checks. This board runs each model in a generic ReAct agent and reports the task pass rate, the percent of tasks whose checks all pass; higher is better.
Metric: Task pass rate (%) on the 1,500 natural-language multi-UAV mission tasks (75 sessions of area search, area assignment and patrol, and target assignment, five difficulty levels) solved through the platform's RESTful APIs and checked by 9,396 hidden validation checks; a task passes when all of its checks pass, with a generic ReAct observe-reason-act agent; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 63.6 | #188 |
| 2 | Seed 2.0 Pro | 30.6 | #211 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=multiuav-plat-react · How It Works · Data refreshed daily, snapshot 2026-11-13.