Can B200 192GB SXM run DeepSeek-V4-Pro?

NVIDIA · 192 GB HBM3e · 8000 GB/s bandwidth · DeepSeek · 861.6B (active 49B) · MoE · model ctx up to 1000K

No — DeepSeek-V4-Pro at Q4 needs ≈ 479.78 GB even at 4K context, beyond the B200 192GB SXM's 192 GB HBM3e. It would only run with heavy CPU/disk offload. Decode at Q4/8K: ≈ 46.6 tok/s (estimated, batch 1). FP16 needs ≈ 1734.0 GB — does not fit on this card. Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.

Fit & speed by quant and context

Quant	Context	VRAM needed	Fits?	tok/s (decode)
Q4	4K	479.78 GB	No	47.5 est.
Q4	8K default	484.68 GB	No	46.6 est.
Q4	32K	514.08 GB	No	41.4 est.
FP16	4K	1729.1 GB	No	—
FP16	8K default	1734.0 GB	No	—
FP16	32K	1763.4 GB	No	—

Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s means the model is far beyond this card (offload-only territory).

VRAM breakdown at 8K context

Quant	Weights	KV cache	Overhead	Total needed	B200 192GB SXM VRAM
Q4	473.88 GB	9.8 GB	1 GB	484.68 GB	192 GB
FP16	1723.2 GB	9.8 GB	1 GB	1734.0 GB	192 GB

Other GPUs that run DeepSeek-V4-Pro

No catalog GPU fully fits DeepSeek-V4-Pro at Q4/8K — see the model page for offload estimates.

All GPUs for DeepSeek-V4-Pro →

Other models for the B200 192GB SXM

• gemma-3-1b-it — ≈5000 tok/s Q4 est.
• Qwen3-0.6B — ≈5000 tok/s Q4 est.
• Llama-3.2-1B-Instruct — ≈4242.4 tok/s Q4 est.
• Qwen3-1.7B — ≈2545.5 tok/s Q4 est.
• Qwen3-30B-A3B — ≈1697.0 tok/s Q4 est.
• Qwen3-Coder-30B-A3B-Instruct — ≈1697.0 tok/s Q4 est.
• GLM-4.7-Flash — ≈1697.0 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈1697.0 tok/s Q4 est.

All models on the B200 192GB SXM →