Key Info
Tencent acknowledged community contributors, including the Maya project, for releasing quantized versions of the 220B Hy4-Flash model, one of which runs on a home PC with an 8GB graphics card.
Highlights
- Maya v1.1.68 ships four self-made quantizations of Hy4-Flash, each benchmarked against the full-precision original
- Maya-XS: 45 GB, 88.5% same top token; Maya-S: 53 GB, 91.8%; Maya-M: 68 GB (further tiers truncated)
- A dedicated variant targets 8GB GPUs, bringing a 220B model to consumer hardware
- Tencent framed the release as part of building alongside the developer community