Mtplx
starred
Application
3x faster speeds on MLX | Qwen 3.8 27B | Native MTP Speculative Decoding On Apple Silicon With No External Drafter.
Plugins Filament, packages Laravel et starter kits open source activement maintenus. Tous en MIT, avec des releases pour v3, v4 et v5 selon les cas.
3x faster speeds on MLX | Qwen 3.8 27B | Native MTP Speculative Decoding On Apple Silicon With No External Drafter.
Giant MoE models on a single consumer GPU by streaming experts from SSD. CUDA fork of antirez/ds4: runs GLM-5.2 (743B), Tencent Hy3 (295B), and DeepSeek 4 Flash, with io_uring expert streaming, LFU host cache, cross-layer expert prefetch, and the first MTP speculative decoding for GLM-5.2.
Tous les packages ont une CI, des tests Pest et des issues ouvertes taguées good first issue.
Nouvelle version disponible.