BBrainOutput

Best hardware to run Qwen3 235B-A22B (MoE)

The minimum self-hostable option and a comfortable recommended build for Qwen3 235B-A22B (MoE), plus the top devices that run it well.

Minimum

NVIDIA B200 (placeholder)

Fits at Q4_K_M (~130GB) with ~39GB headroom — about 1 concurrent instance.

Recommended

NVIDIA B200 (placeholder)

Fits at Q4_K_M (~130GB) with ~39GB headroom — about 1 concurrent instance.

Top devices that run it well

  • NVIDIA B200 (placeholder)
    NVIDIA · Datacenter GPUs

    Fits at Q4_K_M (~130GB) with ~39GB headroom — about 1 concurrent instance.

    Q4_K_M · ~130GBRuns well
  • Supermicro 8x H100 SuperServer
    Supermicro · AI Servers

    Fits at FP16 (~470GB) with ~93.2GB headroom — about 1 concurrent instance.

    FP16 · ~470GBRuns well
  • Dell PowerEdge XE9680
    Dell · AI Servers

    Fits at FP16 (~470GB) with ~93.2GB headroom — about 1 concurrent instance.

    FP16 · ~470GBRuns well
  • AMD Instinct MI300X
    AMD · Datacenter GPUs

    Fits at Q4_K_M (~130GB) with ~39GB headroom — about 1 concurrent instance.

    Q4_K_M · ~130GBRuns well
  • Cloud B200 (Blackwell profile, to verify)
    Cloud · Cloud GPU Profiles

    Fits at Q4_K_M (~130GB) with ~28.4GB headroom — about 1 concurrent instance.

    Q4_K_M · ~130GBRuns well
  • Apple Mac Studio (M2 Ultra)
    Apple · Apple Silicon

    Fits at Q4_K_M (~130GB) but limited bandwidth makes token generation slow for a 235B model.

    Q4_K_M · ~130GBRuns slowly

See all compatible devices →

Run Qwen3 235B-A22B (MoE) on a private AI Business OS

Run your own AI agents on hardware you control — private by design, no per-seat data leaving your premises. BrainOutput helps you pick the right machine and turn it into a working AI Business OS.

Explore the AI Business OS