Can B200 192GB SXM run Kimi-K2-Thinking?

NVIDIA · 192 GB HBM3e · 8000 GB/s bandwidth · Moonshot · 1058.1B (active 32B) · MoE · model ctx up to 256K

No — Kimi-K2-Thinking at Q4 needs ≈ 586.16 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: ≈ 48.2 tok/s (estimated, batch 1). FP16 needs ≈ 2123.6 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	586.16 GB	No	48.8 est.
Q4	8K default	589.36 GB	No	48.2 est.
Q4	32K	608.56 GB	No	45.2 est.
FP16	4K	2120.4 GB	No	—
FP16	8K default	2123.6 GB	No	—
FP16	32K	2142.8 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	581.96 GB	6.4 GB	1 GB	589.36 GB	192 GB
FP16	2116.2 GB	6.4 GB	1 GB	2123.6 GB	192 GB

Other GPUs that run Kimi-K2-Thinking

No catalog GPU fully fits Kimi-K2-Thinking at Q4/8K — see the model page for offload estimates.

All GPUs for Kimi-K2-Thinking →

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 →