Highest Arena reference that fits

Gemma 4 31BRuns great

FP16 · 61.4 GB weights · 63.4 GB total · 36–54 tok/s · Arena reference 1,452

Arena is a published configuration reference. Selected FP16 is untested by this reference.

Arena reference details

Arena Text Overall · style-controlled

1,452FP16 untested

gemma-4-31b · published configuration reference · FP16 not tested

Arena Text Overall (style-controlled), exact published entry gemma-4-31b, 2026-10-08: reference rating 1,452.4. Published confidence bounds 1,444.9–1,459.8. 6,145 votes. Published gemma-4-31b configuration reference. Reasoning mode is not disclosed by the Arena dataset. Tested precision, runtime and sampling settings are not disclosed in this dataset. Selected FP16 has not been measured by this Arena entry. No quantization quality adjustment is applied.

Snapshot published 2026-10-08 · snapshot imported 2026-10-11 UTC · validated last-good data · manual refresh

Arena leaderboard dataset · Arena / lmarena-ai, Arena Leaderboard Dataset (CC BY 4.0). Selected Text Overall style-controlled rows; field names adapted, ratings unchanged. · dataset CC BY 4.0 · revision e16f90232986

Fastest that fits: Qwen3 Coder Next 80B-A3B (MoE) at 273–409 tok/s

Also runs

Open the full calculator →Check another machine

Have a NVIDIA H100 SXM (80GB)? Add your real benchmark → to sharpen these numbers for everyone.

Estimates from device specs and model architectures, refined by community benchmarks. See the methodology.