From 15 tps on llama.cpp to 65 tps with a home-built engine

Open-source inference engine with an OpenAI/Anthropic-style local API that fits the 125B MoE onto a 12GB consumer GPU plus system RAM. The project behind the model's local-running surge.

Posts a video of the 125B MoE running at about 21 tok/s decode with a quarter-million-token window on one gaming card, no quantised cache.

Launch-day announcement of GGUF quants and a guide for CPU and unified-memory setups.

A llama.cpp variant optimised for Ampere cards, reporting very high token rates on 3090 and 3090 Ti, relevant for Flash-Next hosts.