NVIDIA · 80 GB HBM2e · 2039 GB/s bandwidth ·
Google · 30.7B · model ctx up to 256K
Yes — the A100 80GB SXM4 (80 GB HBM2e) can run gemma-4-31B-it. At Q4 with the default 8K context it needs ≈ 24.03 GB VRAM (weights 16.89 GB + KV cache 6.14 GB + 1 GB runtime overhead), which fits within 80 GB. Decode at Q4/8K: ≈ 42.3 tok/s (estimated, batch 1). FP16 (≈ 68.54 GB) also fits, at ≈ 11.6 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
20.96 GB
Yes
42.3est.
Q4
8K default
24.03 GB
Yes
42.3est.
Q4
32K
42.45 GB
Yes
42.3est.
FP16
4K
65.47 GB
Yes
11.6est.
FP16
8K default
68.54 GB
Yes
11.6est.
FP16
32K
86.96 GB
No
28.1est.
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).