Skip to content
Run LLM Locally

Qwen3 235B-A22B on the RTX PRO 6000 Blackwell

Qwen3 235B-A22B at Q4_K_M needs about 141 GB, more than the 96 GB of the RTX PRO 6000 Blackwell: part of the model runs from system RAM, which slows generation. At Q3_K_M it does not fit in VRAM.

ShareFacebookXWhatsAppLinkedInEmail

Qwen3 235B-A22B

Q4_K_M · 8K context

CPU/GPU offload

The model needs about 141 GB but only 93.1 GB of VRAM is usable: 48.3 GB must be loaded into system RAM. It runs, 66% on the GPU, but more slowly.

Your hardware
RTX PRO 6000 Blackwell 96 GB · 64 GB RAM
Estimated memory
~141 GB estimate
VRAM (usable memory)
93.1 GB
GPU only
No
CPU/GPU offload
Required (48.3 GB)
Generation speed
4–7.5 tok/s estimate
Largest context in fast memory
None
Confidence: low

Based on: memory bandwidth (1,792 GB/s) and model size · system RAM bandwidth (~51 GB/s)

Memory breakdown

~141 GB (138 GB–145 GB) needed · estimate

48.3 GB must be offloaded to system RAM

  • Model weights132 GB
  • KV cache1.5 GB
  • Runtime overhead0.8 GB
  • Safety margin6.7 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (CUDA) · flash attention on · weights = real file size

EstimateData updated 24 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card

What makes this estimate less reliable

  • no benchmark for this hardware and model yet
  • split CPU/GPU runs vary a lot with the CPU and PCIe link
  • system RAM speed unknown (dual-channel DDR4-3200 assumed)
  • mixture-of-experts: speed depends on expert routing

Other quantizations on the RTX PRO 6000 Blackwell

Qwen3 235B-A22B on RTX PRO 6000 Blackwell + 64 GB RAM · 8K context
Qwen3 235B-A22B Q8_08K~247 GB · No dataIncompatible
Qwen3 235B-A22B Q6_K8K~191 GB · No dataIncompatible
Qwen3 235B-A22B Q5_K_M8K~166 GB · No dataIncompatible
Qwen3 235B-A22B Q4_K_M8K~141 GB · 4–7.5 tok/s estimateCPU/GPU offload
Qwen3 235B-A22B IQ4_XS8K~125 GB · 5.5–11 tok/s estimateCPU/GPU offload
Qwen3 235B-A22B Q3_K_M8K~112 GB · 8.5–17 tok/s estimateCPU/GPU offload

Alternatives

GPUs that run Qwen3 235B-A22B Q4_K_M entirely in VRAM

    Other models that fit the RTX PRO 6000 Blackwell

    Hardware that runs Qwen3 235B-A22B Q4_K_M entirely in memory

    Prices and links follow the market you choose.

    RTX PRO 6000 Blackwell Workstation Edition

    NVIDIA

    ExcellentRuns Qwen3 235B-A22B Q4_K_M entirely in memory

    Price not available

    Availability unknown

    View on amazon.com

    Affiliate link

    Mac Studio M3 Ultra

    Apple Silicon

    ExcellentRuns Qwen3 235B-A22B Q4_K_M entirely in memory

    Price not available

    Availability unknown

    View on amazon.com

    Affiliate link

    No current offer imported for this market: the link opens a marketplace search.

    Frequently asked questions

    Can the RTX PRO 6000 Blackwell run Qwen3 235B-A22B?
    The model needs about 141 GB but only 93.1 GB of VRAM is usable: 48.3 GB must be loaded into system RAM. It runs, 66% on the GPU, but more slowly.
    Which quantization should I use on the RTX PRO 6000 Blackwell?
    None fits entirely in VRAM at 8K context; the smallest, Q3_K_M, still needs offload.
    What is the largest context on the RTX PRO 6000 Blackwell?
    At Q4_K_M, the model does not fit entirely in VRAM even at a short context.
    How fast is Qwen3 235B-A22B (RTX PRO 6000 Blackwell)?
    We estimate 4–7.5 tok/s (estimate, confidence low). Based on: memory bandwidth (1,792 GB/s) and model size, system RAM bandwidth (~51 GB/s).