Model
Shapelearn Qwen 3.8 27B
A derivative model based on the Qwen series with 27 billion parameters, optimized through quantization. Aggressive quantization reduces VRAM usage to approximately 13.1 GB, enabling local deployment on consumer-grade graphics cards.
Timeline (last 90 days)
Sep 18
A discussion thread about Shapelearn Qwen 3.8 27B appeared on Hacker News, noting a VRAM footprint of 13.1 GB.
Expired50%
Sep 18
Shapelearn's specific technical approach, its improvements over the original Qwen, and the exact meaning of the version number 3.8 are currently difficult to verify.
Unverified50%