DORA (No Tools): leaderboard

DORA (2026) disaster-response tasks answered without tools: a vision-language model sees the raw optical, SAR or multispectral imagery and the question and answers in one step, the tool-free baseline of an agentic benchmark of 515 expert-authored tasks over real disaster sites. Final-answer accuracy in percent, averaged over the five task dimensions; higher is better.

Metric: Final-answer accuracy (%, AVG: unweighted mean of the five task-dimension accuracies) of the tool-free baseline: the vision-language model receives the raw imagery and question and answers in one step without any tool, on the 515 expert-authored DORA disaster-response tasks scored by the same typed rules; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 2 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash18.55#175
2Qwen 3 VL 235B A22B18.3#485

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

Interactive version: theaggregate.ai/benchmark?slug=dora-no-tools · How It Works · Data refreshed daily, snapshot 2026-11-21.