Hugging Face's logo
Full-text search
Search in
Scope to owner or repo
3,873 results

NVFP4 / Polaris-7B-Preview-FP4

README.md

model
1 matches
tags: safetensors, qwen2, nvidia, ModelOpt, Qwen3, quantized, FP4, base_model:deepseek-ai/DeepSeek-R1-Distill-Qwen-7B, base_model:quantized:deepseek-ai/DeepSeek-R1-Distill-Qwen-7B, license:apache-2.0, 8-bit, region:us

NVFP4 / Qwen3-Coder-480B-A35B-Instruct-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-Coder-480B-A35B-Instruct, base_model:quantized:Qwen/Qwen3-Coder-480B-A35B-Instruct, license:apache-2.0, region:us

NVFP4 / DeepSeek-R1-0528-Qwen3-8B-FP4

README.md

model
1 matches
tags: safetensors, qwen3, nvidia, ModelOpt, Qwen3, quantized, FP4, arxiv:2501.12948, base_model:Qwen/Qwen3-8B, base_model:quantized:Qwen/Qwen3-8B, license:mit, 8-bit, region:us

NVFP4 / Qwen3-32B-FP4

README.md

model
1 matches
tags: safetensors, qwen3, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2309.00071, arxiv:2505.09388, base_model:Qwen/Qwen3-32B, base_model:quantized:Qwen/Qwen3-32B, license:apache-2.0, 8-bit, region:us

NVFP4 / DeepSeek-Prover-V2-7B-FP4

README.md

model
1 matches
tags: safetensors, llama, nvidia, ModelOpt, Qwen3, quantized, FP4, base_model:deepseek-ai/DeepSeek-Prover-V2-7B, base_model:quantized:deepseek-ai/DeepSeek-Prover-V2-7B, license:mit, 8-bit, region:us

NVFP4 / Polaris-4B-Preview-FP4

README.md

model
1 matches
tags: safetensors, qwen3, nvidia, ModelOpt, Qwen3, quantized, FP4, en, base_model:Qwen/Qwen3-4B, base_model:quantized:Qwen/Qwen3-4B, license:apache-2.0, 8-bit, region:us

NVFP4 / Qwen3-235B-A22B-Instruct-2507-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-235B-A22B-Instruct-2507, base_model:quantized:Qwen/Qwen3-235B-A22B-Instruct-2507, license:apache-2.0, region:us

NVFP4 / Qwen3-235B-A22B-Thinking-2507-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-235B-A22B-Thinking-2507, base_model:quantized:Qwen/Qwen3-235B-A22B-Thinking-2507, license:apache-2.0, 8-bit, region:us

NVFP4 / Qwen3-30B-A3B-Thinking-2507-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-30B-A3B-Thinking-2507, base_model:quantized:Qwen/Qwen3-30B-A3B-Thinking-2507, license:apache-2.0, 8-bit, region:us

NVFP4 / Qwen3-30B-A3B-Instruct-2507-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-30B-A3B-Instruct-2507, base_model:quantized:Qwen/Qwen3-30B-A3B-Instruct-2507, license:apache-2.0, 8-bit, modelopt, region:us

NVFP4 / Qwen3-Coder-30B-A3B-Instruct-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3_moe, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-Coder-30B-A3B-Instruct, base_model:quantized:Qwen/Qwen3-Coder-30B-A3B-Instruct, license:apache-2.0, 8-bit, modelopt, region:us

NVFP4 / Qwen3-0.6B-FP4

README.md

model
1 matches
tags: Model Optimizer, safetensors, qwen3, nvidia, ModelOpt, Qwen3, quantized, FP4, text-generation, conversational, arxiv:2505.09388, base_model:Qwen/Qwen3-0.6B, base_model:quantized:Qwen/Qwen3-0.6B, license:apache-2.0, 8-bit, region:us

NVFP4 / LTX2.3-NVFP4-Diffusers

model
1 matches
tags: safetensors, region:us

Henley04 / SoulX-Singer-nvfp4

README.md

model
44 matches
tags: nvfp4, SoulX-Singer, svs, text-to-speech, zh, en, arxiv:2602.07803, base_model:Soul-AILab/SoulX-Singer, base_model:finetune:Soul-AILab/SoulX-Singer, license:apache-2.0, region:us
14
# SoulX-Singer NVFP4 (Quantized) 
15
 
16
This is the **NVIDIA FP4 (NVFP4)** weight-only quantized version of the original 
⋯
19
It is produced by a one-time `torchao` NVFP4 quantization pass over the official 
⋯
29
- **Tooling:** `torchao>=0.17.0` (`NVFP4WeightOnlyConfig`) 
⋯
33
## Why NVFP4 
34
 
35
NVFP4 is the **native 4-bit floating-point format** introduced with the 
36
Blackwell architecture. Compared with INT4 / W4A16 pseudo-quantization, 
37
NVFP4 has dedicated hardware units on sm100+ and runs the **actual 4-bit 
⋯
45
quantized to NVFP4. The following tensors stay in their original precision: 

sroecker / Qwen3.6-35B-REAP-Pruned-ratio-0.5-NVFP4

README.md

model
6 matches
tags: safetensors, qwen3_5_moe, qwen, nvfp4, vllm, compressed-tensors, base_model:RangerX/Qwen3.6-35B-REAP-Pruned-ratio-0.5, base_model:quantized:RangerX/Qwen3.6-35B-REAP-Pruned-ratio-0.5, 8-bit, region:us
13
# NVFP4 Quantized Qwen3.6-35B-REAP-Pruned-ratio-0.5-NVFP4 
14
 
15
This is an NVFP4 quantized version of 
16
`RangerX/Qwen3.6-35B-REAP-Pruned-ratio-0.5` with `llm-compressor`. 
17
The model has both weights and activations quantized to NVFP4 format in 
⋯
26
Tested on an NVIDIA GeForce RTX 5070 Ti 16 GiB with vLLM's NVFP4 linear path 
⋯
164
scheme="NVFP4", 

prithivMLmods / Q3.6-27B-GLM-5.1-DA-nvfp4

README.md

model
2 matches
tags: transformers, safetensors, qwen3_5, image-text-to-text, text-generation-inference, vllm, glm-5.1, math, reasoning, pytorch, uncensored, abliterated, unfiltered, unredacted, refusal-ablated, alignment-modified, nvfp4, conversational, en, dataset:prithivMLmods/harm_bench, dataset:Jackrong/GLM-5.1-Reasoning-1M-Cleaned, base_model:prithivMLmods/Q3.6-27B-GLM-5.1-DA, base_model:quantized:prithivMLmods/Q3.6-27B-GLM-5.1-DA, license:apache-2.0, endpoints_compatible, compressed-tensors, region:us
28
# **NVFP4 — Q3.6-27B-GLM-5.1-DA** 
29
 
30
> bf16 : https://huggingface.co/prithivMLmods/Q3.6-27B-GLM-5.1-DA 

dryade36513 / 10Eros_v1_experimental_nvfp4

README.md

model
2 matches
tags: nvfp4, base_model:Lightricks/LTX-2.3, base_model:finetune:Lightricks/LTX-2.3, license:mit, region:us
9
NVFP4 conversion of https://huggingface.co/TenStrip/LTX2.3-10Eros/tree/main 
10
Transformer only, needs distill lora. 
11
 
12
Used https://github.com/silveroxides/convert_to_quant 

ApacheOne / ZImageTurbo-nvfp4_mixed

README.md

model
3 matches
tags: custom, quantization, nvfp4, art, en, base_model:Tongyi-MAI/Z-Image-Turbo, base_model:quantized:Tongyi-MAI/Z-Image-Turbo, license:apache-2.0, region:us
14
## nvfp4 model infomation 
15
Note: This model has not been personally tested but should work in ComfyUI. Feedback is appreciated. 
⋯
30
Note: Custom FP32 mixed nvfp4 version of base ZImageturbo  

ApacheOne / ZImageBase-nvfp4_mixed

README.md

model
3 matches
tags: custom, quantization, nvfp4, art, en, base_model:notaneimu/z-image-base-comfy-fp32, base_model:quantized:notaneimu/z-image-base-comfy-fp32, license:apache-2.0, region:us
14
## nvfp4 model infomation 
⋯
31
Note: Custom FP32 mixed nvfp4 version of base ZImagebase 

ItBitter / ZImageTurbo-nvfp4_mixed

README.md

model
3 matches
tags: custom, quantization, nvfp4, art, en, base_model:Tongyi-MAI/Z-Image-Turbo, base_model:quantized:Tongyi-MAI/Z-Image-Turbo, license:apache-2.0, region:us
14
## nvfp4 model infomation 
15
Note: This model has not been personally tested but should work in ComfyUI. Feedback is appreciated. 
⋯
30
Note: Custom FP32 mixed nvfp4 version of base ZImageturbo