---
format: "aidr-story-markdown/v1"
id: "37157a69a1755c3be41ce17688119fcfb2bdf843b024f7ac74e1b5bf6e8f4361"
canonical_url: "https://aidr.today/37157a69?lang=en"
title: "Open Source 4B VLM Beats Qwen 122B in GeoGuesser Benchmarks"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-09T19:18:06.000Z"
category: "Research"
topics: ["vision-language","open-source","qwen","benchmark","reasoning"]
source_urls: ["https://huggingnews.com/ai/open-source-4b-vlm-beats-qwen-122b-in-geoguesser-benchmarks-8acc02fa"]
summary: "Reinforcement learning was used to train a small scale vision model to identify locations on digital maps. The 4 billion parameter VLM beat Qwen 3.5 122B and G…"
---

# Open Source 4B VLM Beats Qwen 122B in GeoGuesser Benchmarks

> [Open the canonical story](<https://aidr.today/37157a69?lang=en>)

**Published:** 2026-09-09T19:18:06.000Z
**Category:** Research
**Topics:** vision\-language, open\-source, qwen, benchmark, reasoning

## Summary

Reinforcement learning was used to train a small scale vision model to identify locations on digital maps\. The 4 billion parameter VLM beat Qwen 3\.5 122B and G…

## Sources

- [Story source](<https://huggingnews.com/ai/open-source-4b-vlm-beats-qwen-122b-in-geoguesser-benchmarks-8acc02fa>)