artificialbuilds

Strata: one-click engine for running the 125B Flash-Next on a gaming PC

NikoSep 2416.8K stars

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.

An 8-second preview. The full video is on GitHub.

Launch: Qwen3.8-Flash-Next
Use case: Local inference
Format: Tool
Source: github.com

Niko

AI & fun

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

Tells the story of the 12GB RTX 5070 user who wrote Strata after finding llama.cpp too slow.

Post · 1.6K likes

Flash-Next on a 24GB RTX 4090 with 250K context

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.

Build · 4.7K likes

Unsloth GGUFs: run Flash-Next on 75GB RAM

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

Post · 2.5K likes

llamAmpere: llama.cpp fork tuned for RTX 3090

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

Tool · 148 stars