Can I run Qwen3-8B-MLX-bf16 on an NVIDIA H200?

Yes

Its published weights (16.4 GB) fit in the 127 GB an NVIDIA H200 can use, with room for ≈ 40K tokens of context.

Full analysis of Qwen3-8B-MLX-bf16 Everything that runs on NVIDIA H200 Get the weights on Hugging Face

hf download Qwen/Qwen3-8B-MLX-bf16

Files and context on NVIDIA H200

Each file of Qwen3-8B-MLX-bf16 (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.

File	Size	On NVIDIA H200	Context (16-bit)	Context (8-bit)
As published	16.4 GB	Fits	40K	40K
4-bit (estimate)	4.61 GB	Fits	40K	40K

Qwen3-8B-MLX-bf16 on other hardware

NVIDIA H200Yes
NVIDIA H100Yes
RTX 5090Yes
RTX 4090Yes
RTX 3090Yes
RTX 5060 Ti 16 GBProbably, at 4-bit
RTX 3060 12 GBProbably, at 4-bit
RTX 4060Probably, at 4-bit
Mac with 32 GBYes
Mac with 64 GBYes
Mac with 128 GBYes

Popular models that run on NVIDIA H200

Qwen3-0.6BQwen · 752M
Qwen3-VL-8B-InstructQwen · 8.77B
gpt2openai-community · 137M
Qwen3-8BQwen · 8.19B
Qwen3-Coder-30B-A3B-Instruct-GGUFunsloth · 30.5B
gemma-4-26B-A4B-itgoogle · 25.8B
Qwen2.5-7B-InstructQwen · 7.62B
Qwen3.5-9BQwen · 9.65B

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.