Model Anatomy See what's inside open-weight AI models

Can I run QwQ-32B-GGUF on an NVIDIA H200?

Yes

Its qwq-32b-fp16 file (65.5 GB) fits in the 127 GB an NVIDIA H200 can use, with room for ≈ 128K tokens of context.

Full analysis of QwQ-32B-GGUF Everything that runs on NVIDIA H200 Get the qwq-32b-fp16 file on Hugging Face
llama-server -hf Qwen/QwQ-32B-GGUF

Files and context on NVIDIA H200

Each file of QwQ-32B-GGUF (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.

FileSizeOn NVIDIA H200 Context (16-bit)Context (8-bit)
qwq-32b-fp1665.5 GBFits128K128K
Q8_034.8 GBFits128K128K
Q6_K26.9 GBFits128K128K
Q5_K_M23.3 GBFits128K128K
Q5_022.6 GBFits128K128K
Q4_K_M19.9 GBFits128K128K

QwQ-32B-GGUF on other hardware

Popular models that run on NVIDIA H200

Assumes 90% of the card's memory is usable and at least 4K tokens of context for one conversation, with the context cache in 16-bit unless noted. Fits leaves at least 20% of that memory for context; Tight leaves less.