gemma-4-31B-it-qat-NVFP4-Blackwell
> [!TIP] > There is a newer version of this model: > gemma-4-31B-it-qat-NVFP4-mse-Blackwell. > Same base model and calibration set, but with MSE-based weight scaling > instead of absmax, which keeps outliers from crushing the rest of each block. > This repo stays as it is so existing deployments don't change under their feet.
Params
31.0 B
Context
—
Downloads 30d
371 K
Likes
4
Download history
daily snapshots · 9 days371 K177 K
Oct 16Oct 19Oct 22Oct 24
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 | 20.4 GB | 27.6 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 -s https://aimodelscomparison.com/api/v1/models/gemma-4-31b-it-qat-nvfp4-blackwell
{
"hf_id": "melcheikh/gemma-4-31B-it-qat-NVFP4-Blackwell",
"params_b": 31.00,
"context_length": null,
"license": { "id": "apache-2.0", "commercial": "yes" },
"downloads_30d": 371496,
"vram_estimates": [
{ "quant": "u8", "gb": 27.6 }
],
"updated_at": "2026-10-16T01:00:32Z"
}
Specifications
- Architecture
- Gemma4ForConditionalGeneration
- Parameters
- 31.0 B
- Tensor type
- U8
- Licence
- apache-2.0
- First seen on the Hub
- 2026-06-05
- Training datasets
- undisclosed
- Added to our catalog
- 2026-10-16
Compare with any text-generation model
Popular comparisons
Appears in