← AI Hashrate

Can H200 141GB SXM5 run gemma-3-12b-it?

NVIDIA · 141 GB HBM3e · 4800 GB/s bandwidth · Google · 12.2B · model ctx up to 128K

Yes — the H200 141GB SXM5 (141 GB HBM3e) can run gemma-3-12b-it. At Q4 with the default 8K context it needs ≈ 10.11 GB VRAM (weights 6.71 GB + KV cache 2.4 GB + 1 GB runtime overhead), which fits within 141 GB. Decode at Q4/8K: ≈ 254.5 tok/s (estimated, batch 1). FP16 (≈ 27.8 GB) also fits, at ≈ 70.0 tok/s (estimated). 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

QuantContextVRAM neededFits?tok/s (decode)
Q44K8.91 GBYes254.5 est.
Q48K default10.11 GBYes254.5 est.
Q432K17.31 GBYes254.5 est.
FP164K26.6 GBYes70.0 est.
FP168K default27.8 GBYes70.0 est.
FP1632K35.0 GBYes70.0 est.

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

QuantWeightsKV cacheOverheadTotal neededH200 141GB SXM5 VRAM
Q46.71 GB2.4 GB1 GB10.11 GB141 GB
FP1624.4 GB2.4 GB1 GB27.8 GB141 GB

Other GPUs that run gemma-3-12b-it

All GPUs for gemma-3-12b-it →

Other models for the H200 141GB SXM5

All models on the H200 141GB SXM5 →