MultiGlobeQA (Web Search Agent): leaderboard
MultiGlobeQA (2026) is a multilingual geospatial reasoning benchmark of execution-verified questions over three knowledge graphs, covering 14 spatial-function families and 201 countries. Exact-match accuracy of a code agent that retrieves by web search, in percent; higher is better.
Metric: Exact-match accuracy (%; Tier 3b: a CodeAgent writing and executing Python with web-search tools, at most 10 steps; exact match with answer-type tolerances (geodesic or relative tolerance, set equality, polygon IoU of at least 0.5) over the 4,979 true-premise questions of the English text small split, refusals and non-answers counted incorrect; greedy decoding, output constrained to the answer-type JSON schema; mean of three seeds). Source: arxiv.org. Saturation forecast: Around 2029. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 27.2 | #141 |
| 2 | Qwen 3.5 35B A3B FP8 | 22.8 | #488 |
| 3 | Qwen 3.5 27B FP8 | 22.6 | #269 |
| 4 | Gemma 3 27B (IT) | 21.7 | #1306 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=multiglobeqa-web-search-agent · How It Works · Data refreshed daily, snapshot 2026-11-12.