Can I run Qwen3-Next-80B-A3B-Thinking on an NVIDIA B200?

Yes, with a GGUF version

The published weights are too big, but the Qwen3-Next-80B-A3B-Thinking file from Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF (159 GB) fits, with room for ≈ 98K tokens of context.

Full analysis of Qwen3-Next-80B-A3B-Thinking Everything that runs on NVIDIA B200 Get the Qwen3-Next-80B-A3B-Thinking file on Hugging Face

llama-server -hf Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF

Files and context on NVIDIA B200

Each file of Qwen3-Next-80B-A3B-Thinking (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 B200	Context (16-bit)	Context (8-bit)
As published	163 GB	Doesn't fit	—	—
Qwen3-Next-80B-A3B-ThinkingQwen/Qwen3-Next-80B-A3B-Thinking-GGUF	159 GB	Tight	98K	196K
Q8_0Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF	84.8 GB	Fits	256K	256K
Q6_KQwen/Qwen3-Next-80B-A3B-Thinking-GGUF	65.5 GB	Fits	256K	256K
Q5_K_MQwen/Qwen3-Next-80B-A3B-Thinking-GGUF	56.7 GB	Fits	256K	256K
Q5_0Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF	55.0 GB	Fits	256K	256K
Q4_K_MQwen/Qwen3-Next-80B-A3B-Thinking-GGUF	48.4 GB	Fits	256K	256K

Qwen3-Next-80B-A3B-Thinking on other hardware

Popular models that run on NVIDIA B200

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.