GeoThinker-F Study - Scaled - VSI-Bench: leaderboard
S3.T2; new GeoThinker-F trained checkpoints only; Scaled training regime retained as distinct identity; quoted baseline literature scores omitted; candidate coverage is not fitted eligibility. Metric: Accuracy (%). Higher is better. Source: https://arxiv.org/html/2610.20592v1
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GeoThinker-F-Qwen3VL-8B (Scaled) | 70.4 | |
| 2 | GeoThinker-F-Qwen2.5VL-7B (Scaled) | 67.3 |
Interactive version: theaggregate.ai/benchmark?slug=geothinker-f-study-scaled-vsi-bench · How It Works · Data refreshed daily, snapshot 2026-10-16.