THUSI-Lab / GameScaling
git clone https://github.com/THUSI-Lab/GameScaling.git
Inference client and checkpoints for GameScaling, vision-language models that predict keyboard and mouse actions from game frames.
Why you might care
Provides Qwen3.5-based VLMs continued-pretrained on 1,600 hours of keyboard-and-mouse gameplay, in 0.8B to 27B sizes. Each call takes recent frames and actions and returns the next 200 ms of inputs; the repo contains only the serving client and prompt template, with no training code.
+9
stars, 24h · #4,315
+48
stars, 7 days · #6,060
+494
stars, 30 days · #2,190
494
★ total · +11% this week
28
forks · 2 open issues
Star history (as recorded here)
LanguagePython
LicenseApache-2.0
Created2026-10-10 (27d ago)
Last push8d ago
Watchers1
OwnerOrganization
GitHub topics
Links
- GitHub: github.com/THUSI-Lab/GameScaling
- Homepage: https://huggingface.co/THUSI-Lab/GameScaling
- Releases · Issues · Hacker News mentions
README
Loading README…
Blurb generated by claude-haiku-5-5 11d ago. Star deltas marked ~ are extrapolated from partial history.