NVIDIA · 192 GB HBM3e · 8000 GB/s bandwidth ·
Mistral · 24.0B · model ctx up to 128K
Yes — the B200 192GB SXM (192 GB HBM3e) can run Magistral-Small-2509. At Q4 with the default 8K context it needs ≈ 19.0 GB VRAM (weights 13.2 GB + KV cache 4.8 GB + 1 GB runtime overhead), which fits within 192 GB. Decode at Q4/8K: ≈ 212.1 tok/s (estimated, batch 1). FP16 (≈ 53.8 GB) also fits, at ≈ 58.3 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
Quant
Context
VRAM needed
Fits?
tok/s (decode)
Q4
4K
16.6 GB
Yes
212.1est.
Q4
8K default
19.0 GB
Yes
212.1est.
Q4
32K
33.4 GB
Yes
212.1est.
FP16
4K
51.4 GB
Yes
58.3est.
FP16
8K default
53.8 GB
Yes
58.3est.
FP16
32K
68.2 GB
Yes
58.3est.
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).