fraQtl/Qwen3.6-35B-A3B-Hi-Fi-GGUF

Text Generation • 35B • Updated 10 days ago • 942 • 4

Note ≈23% lower KLD than the leading public Q4_K_M at identical file size — MoE-aware per-tensor policy

fraQtl/Qwen3.6-35B-A3B-Hi-Fi-MTP-runtime

Text Generation • 36B • Updated May 31 • 150 • 2

Note Datacenter variant: 1.49× decode via MTP speculative decoding (A100-80GB)

fraQtl/Gemma-4-E2B-it-Hi-Fi-GGUF

Text Generation • 5B • Updated 10 days ago • 1.64k • 13

Note 36% closer to the original than the community Q4_K_M at identical size — plus a 17.5%-smaller phone artifact

fraQtl/Gemma-4-26B-A4B-it-Hi-Fi-GGUF

Text Generation • 25B • Updated 10 days ago • 326

Note −9% KLD to the original at same size — with the GSM8K loss disclosed up front (honest split)