NPU-first runtime

Run frontier LLMs directly on Ryzen™ AI NPUs

FastFlowLM delivers an Ollama-style developer experience that is tuned for AMD’s XDNA2 accelerators: install in seconds, stream tokens instantly, and sustain up to 256k context windows with 10× better efficiency than GPU-first stacks.

Download FastFlowLM GitHub Benchmarks Models Demos Test Drive Discord

  • Runtime size

    ~16 MB

  • Context

    Up to 256k tokens

  • Supported chips

    Ryzen™ AI (Strix, Strix Halo, Kraken)

fastflow.ps1 
Invoke-WebRequest https://github.com/FastFlowLM/FastFlowLM/releases/latest/download/flm-setup.exe `
  -OutFile flm-setup.exe
Start-Process .\flm-setup.exe -Wait

flm pull llama3.2:1b
flm run llama3.2:1b
flm serve gemma3:4b --vision
LLMs + VLMs Embeddings Audio + Whisper Agentic APIs