Downloads · 30 days
9
9% of all-time downloads
gpt-oss-20b-NVFP4
arathishree/gpt-oss-20b-NVFP4
gpt-oss-20b-NVFP4 is a machine learning model from arathishree. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
GPT-OSS-20B NVFP4 Blockwise NVFP4 quantization of gpt-oss-20b using NVIDIA ModelOpt. Quantization Details
All-time downloads
97
Public
Parameters
20.6B
81.9 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors40.9 GB · 100%
Parameter types
How the weights are stored.
BF1620.3B · 98%
Base models
Model card
From the Hugging Face model README
GPT-OSS-20B NVFP4 Blockwise NVFP4 quantization of gpt-oss-20b using NVIDIA ModelOpt. Quantization Details
Method: Blockwise GPTQ (NVFP4) Calibration: WikiText-2, 512 samples Layers Quantized: 1,342/1,632 (82.2%) Quantization Time: 55.8 minutes
Quality Metrics DatasetBF16 pplNVFP4 pplΔpplWikiText-2226.79220.29-6.50C4708.94574.88-134.06
---
license: apache-2.0
base_model: gpt-oss-20b
tags:
- quantization
- nvfp4
- gptq
- moe
---
GPT-OSS-20B NVFP4
Blockwise NVFP4 quantization of gpt-oss-20b using NVIDIA ModelOpt.
Quantization Details
- Method: Blockwise GPTQ (NVFP4)
- Calibration: WikiText-2, 512 samples
- Layers Quantized: 1,342/1,632 (82.2%)
- Quantization Time: 55.8 minutes
Quality Metrics
| Dataset | BF16 ppl | NVFP4 ppl | Δppl |
|---|---|---|---|
| WikiText-2 | 226.79 | 220.29 | -6.50 |
| C4 | 708.94 | 574.88 | -134.06 |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model = AutoModelForCausalLM.from_pretrained(
"arathishree/gpt-oss-20b-NVFP4",
dtype=torch.bfloat16,
device_map="auto",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("arathishree/gpt-oss-20b-NVFP4", trust_remote_code=True)
prompt = "The capital of France is"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=80)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
- Type
- gpt_oss
- License
- apache-2.0
- Created
- Jul 14, 2026
- Updated
- Jul 14, 2026