EVGeoQA (Hangzhou): leaderboard

Dual-objective geo-spatial exploration on EVGeoQA (2026), Hangzhou split: from the user's live coordinate the agent must find an electric-vehicle charging station that is also within walking distance of a place for a second activity, using only local search, move and distance tools on real map data. No Limit tier (any distance to the target); Hits@1 in percent, higher is better.

Metric: Hits@1 (%, times 100) in the No Limit tier (no cap on the distance from the user to the target station) for Hangzhou queries of EVGeoQA: the top recommended charging station must be in the ground-truth set of up to five stations that are walkable to a POI matching the user's second intent, found by the agent through the GeoRover tools (5 km station search, 1 km POI search, location moves, driving distance) on Gaode map data; higher is better. Source: arxiv.org. Saturation forecast: Around March 2028. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 2.5 Pro36.16#276
2Gemini 2.5 FlashNon-reasoning35.13#463
3Qwen 2.5 72B33.33#1188
4Qwen 3 8B26.37#1873
5GPT-OSS-20B26.12#1183
6Qwen 3 30B A3BNon-reasoning23.39#1156
7GPT-OSS-120B18.42#704

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

Interactive version: theaggregate.ai/benchmark?slug=evgeoqa-hangzhou · How It Works · Data refreshed daily, snapshot 2026-11-21.