← AI Hashrate

Can A100 80GB SXM4 run Qwen3-14B?

NVIDIA · 80 GB HBM2e · 2039 GB/s bandwidth · Alibaba · 14.8B · model ctx up to 128K

Yes — the A100 80GB SXM4 (80 GB HBM2e) can run Qwen3-14B. At Q4 with the default 8K context it needs ≈ 12.1 GB VRAM (weights 8.14 GB + KV cache 2.96 GB + 1 GB runtime overhead), which fits within 80 GB. Decode at Q4/8K: ≈ 87.7 tok/s (estimated, batch 1). FP16 (≈ 33.56 GB) also fits, at ≈ 24.1 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)
Q44K10.62 GBYes87.7 est.
Q48K default12.1 GBYes87.7 est.
Q432K20.98 GBYes87.7 est.
FP164K32.08 GBYes24.1 est.
FP168K default33.56 GBYes24.1 est.
FP1632K42.44 GBYes24.1 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 neededA100 80GB SXM4 VRAM
Q48.14 GB2.96 GB1 GB12.1 GB80 GB
FP1629.6 GB2.96 GB1 GB33.56 GB80 GB

Other GPUs that run Qwen3-14B

All GPUs for Qwen3-14B →

Other models for the A100 80GB SXM4

All models on the A100 80GB SXM4 →