Can RTX 5080 16GB run gemma-4-31B-it-FP8-block?

NVIDIA · 16 GB GDDR7 · 960 GB/s bandwidth · RedHatAI · 31.3B · model ctx up to NoneK

Fit & speed by quant and context

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

Other GPUs that run gemma-4-31B-it-FP8-block

• B200 192GB SXM — ≈277.4 tok/s Q4 est.
• MI300X 192GB — ≈173.8 tok/s Q4 est.
• H200 141GB SXM5 — ≈166.4 tok/s Q4 est.
• H100 80GB SXM5 — ≈116.2 tok/s Q4 est.
• Gaudi 3 128GB — ≈107.4 tok/s Q4 est.
• A100 80GB SXM4 — ≈70.7 tok/s Q4 est.
• RTX 5090 32GB — ≈62.1 tok/s Q4 est.
• RTX 6000 Ada 48GB — ≈33.0 tok/s Q4 est.

All GPUs for gemma-4-31B-it-FP8-block →

Other models for the RTX 5080 16GB

• MiniCPM5-1B — ≈955.3 tok/s Q4 est.
• Qwen3.5-2B — ≈515.9 tok/s Q4 est.
• DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx — ≈448.6 tok/s Q4 est.
• Llama-3.2-3B-Instruct — ≈322.4 tok/s Q4 est.
• gemma-4-26B-A4B-it — ≈271.5 tok/s Q4 est.
• Qwen3-4B-Instruct-2507 — ≈257.9 tok/s Q4 est.
• NVIDIA-Nemotron-3-Nano-4B-BF16 — ≈257.9 tok/s Q4 est.
• Qwen3-4B-Thinking-2507 — ≈257.9 tok/s Q4 est.

All models on the RTX 5080 16GB →