LilaRest / text-generation updated 2 weeks ago

gemma-4-31B-it-NVFP4-turbo

Name: gemma-4-31B-it-NVFP4-turbo
Author: LilaRest

A repackaged nvidia/Gemma-4-31B-IT-NVFP4 that is 68% smaller in GPU memory and ~2.5× faster than the base model, while retaining nearly identical quality (1-3% loss). Fits on a single RTX 5090 (🎉).

Params

32.5 B

Context

—

Downloads 30d

702 K

Likes

303

Commercial use: allowed apache-2.0 Not gated SAFETENSORS View on Hugging Face ↗

Download history

daily snapshots · 10 days

705 K608 K

Jul 28Jul 31Aug 3Aug 6

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	u8	19.3 GB	26.6 GB	❌ Won’t fit

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub

~ · curl · api/v1

$ curl -s https://aimodelscomparison.com/api/v1/models/gemma-4-31b-it-nvfp4-turbo
{
  "hf_id": "LilaRest/gemma-4-31B-it-NVFP4-turbo",
  "params_b": 32.53,
  "context_length": null,
  "license": { "id": "apache-2.0", "commercial": "yes" },
  "downloads_30d": 702267,
  "vram_estimates": [
    { "quant": "u8", "gb": 26.6 }
  ],
  "updated_at": "2026-07-28T18:04:42Z"
}

est. VRAM —on RTX 4090 · 24 GBJSON API →

Specifications

Architecture: Gemma4ForCausalLM
Parameters: 32.5 B
Tensor type: U8
Licence: apache-2.0
First seen on the Hub: 2026-04-07
Base model: gemma-4-31B-it
Training datasets: undisclosed
MMLU Pro (reported): 83.93
GPQA Diamond (reported): 72.73
Added to our catalog: 2026-07-28

Family

Base model and the most-downloaded derivatives in the catalog.

Base gemma-4-31B-it google 11.9 M

Compare with

Qwen3-8B 16.3 M Qwen3.6-35B-A3B-NVFP4 11.3 M gpt-oss-20b 8.6 M Qwen3-32B 8.3 M dolphin-2.9.1-yi-1.5-34b 4.6 M OTel-2.0-LLM-31B-IT 3.3 M

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB