Tencent's 220B Hy4-flash quantized to 49 GB weights, runnable on 64 GB VRAM
gladkos · reddit · 2026-11-19
A community GGUF build of Tencent's 220B MoE model Hy4-flash shrinks weights from 70 GB to 49.1 GB, fitting a 64 GB VRAM machine.
- Architecture: 220B total = 155.5B network (512 routed experts, 16 active per token, 11.3B active parameters) + 63.5B Engram, a hash-lookup table of learned bigram/trigram embeddings; text and image input, 1M-token context.
- Quantization scheme: experts at 2.3 bits, everything else Q80; the Engram table lives in its own file read a few KB per token, so it stays on SSD.
- Comparison: Tencent's own AngelSlim GGUF builds keep 72–105 GB in memory, better suited to 96/128 GB machines; this single-file build targets 64 GB boxes.
More from Infra
- tinygrad lists $89,000 tinybox with 768GB VRAM: four refurbished AMD MI300X GPUs — viksit · 2026-11-19
- RustingStrata vs Strata on RTX 3060: Rust MoE engine edges out 28.3 vs 26.1 tok/s — Serious_Scarcity7818 · 2026-11-19
- Investor: market hates uncertainty more than money after Nvidia earnings — MartinGTobias · 2026-11-19
- SGLang ships Day-0 serving support for new model DaisyRose-27B — ying11231 · 2026-11-19
- Linux LTS patch flood: 3318 patches for 6.6 alone, 'no way humans are reviewing this' — pkqzy888 · 2026-11-19
- DaisyRose-27B's in-house sandbox: ~80% cheaper than cloud providers, 99.95% success rate — BanghuaZ · 2026-11-19