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

Can I run Qwen2.5-7B-Instruct-GGUF on an NVIDIA H200?

Yes

Its qwen2.5-7b-instruct-fp16 file (15.2 GB) fits in the 127 GB an NVIDIA H200 can use, with room for ≈ 128K tokens of context.

Full analysis of Qwen2.5-7B-Instruct-GGUF Everything that runs on NVIDIA H200 Get the qwen2.5-7b-instruct-fp16 file on Hugging Face
llama-server -hf Qwen/Qwen2.5-7B-Instruct-GGUF

Files and context on NVIDIA H200

Each file of Qwen2.5-7B-Instruct-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)
qwen2.5-7b-instruct-fp1615.2 GBFits128K128K
Q8_08.10 GBFits128K128K
Q6_K6.25 GBFits128K128K
Q5_K_M5.44 GBFits128K128K
Q5_05.32 GBFits128K128K
Q4_K_M4.68 GBFits128K128K
Q4_04.43 GBFits128K128K
Q3_K_M3.81 GBFits128K128K
Q2_K3.02 GBFits128K128K

Qwen2.5-7B-Instruct-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.