GeoBrowse - Level 2 City (GATE Agent): leaderboard

GeoBrowse (2026) Level 2: models geolocate an image and then follow an obfuscated BrowseComp-style multi-hop knowledge query. This board scores the city-level targets with the model inside the GATE agent workflow of the benchmark with image and web tools; accuracy in percent from an LLM judge, mean of three runs, higher is better.

Metric: Accuracy (%) on the city-level targets of GeoBrowse Level 2 (101 BrowseComp-style instances that add long-tail knowledge and obfuscated entities on top of the geolocation; 19 city, 19 state and 63 country targets), judged correct by an LLM judge with the Humanity's Last Exam judging prompt, mean of 3 runs, under the benchmark's GATE agentic workflow: a ReAct loop with five image tools (crop, super-resolution and others) and four knowledge tools (web text and image search, page visits, code) under a fixed protocol, tool suite and call budget; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro26.3#146
2GPT-521.1#185
3Claude Opus 4.521.1#165
4GPT-4o15.8#754
5Gemini 2.5 Pro15.8#280
6GPT-4.115.8#506
7Llama 3.2 90B5.3#1761
8Qwen 3 VL 32B5.3#624
9Qwen 2.5 VL 7B0#1878

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=geobrowse-level-2-city-gate-agent · How It Works · Data refreshed daily, snapshot 2026-11-22.