AnalysisAI ModelsOctober 16, 2026

capy-flash runs Qwen3.8-Flash-Next 125B MoE in NVFP4 on two 16GB GPUs

Read original source →reddit.com

capy-flash, an evolution of the ninfer-tp2 engine, runs the 125B-parameter Qwen3.8-Flash-Next MoE in full NVFP4 across two 16 GB consumer GPUs. Author credits Strata as inspiration, noting Strata added NVFP4 support mid-build.

1 source

More stories today

Open the live feed