Qwen3.8-Flash-Next NVFP4
Qwen3.8-Flash-Next NVFP4 is a mixed-precision, post-training-quantized checkpoint of the Qwen3.8-Flash-Next multimodal causal language model prepared with NVIDIA Model Optimizer. It provides a smaller, GPU-accelerated checkpoint intended for deployment with vLLM on NVIDIA Blackwell GPUs and supports long-context multimodal (text, image, video) inputs with text outputs.
What it does
Developers can run it locally for AI agents, chatbots, RAG systems, and other AI-powered applications.
- Availability
- Released on Hugging Face 10/05/2026
- Pricing model
- free
- Deployment
- Local deployment; validated on NVIDIA DGX Spark and NVIDIA RTX PRO 6000 Blackwell
- Privacy
- governed by NVIDIA Open Model License; refers to Qwen Community License 1.0
- Licence
- NVIDIA Open Model Agreement; base model governed by Qwen Community License 1.0
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
12 claims- The repository contains model weights and configuration files for the post-trained model. [1]
> This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, and TokenSpeed. [1]
> These artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, TokenSpeed, etc.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - The NVFP4 1.0 checkpoint was released on Hugging Face on 08/31/2026. [2]
The model version is NVFP4 1.0 and was produced with nvidia-modelopt **v0.46.0**.
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model supports vLLM and is listed as compatible with NVIDIA Blackwell B200 and B300. [2]
- The NVFP4 checkpoint is approximately 2.7 times smaller than the BF16 source checkpoint. [2]
The resulting checkpoint is approximately 2.7x smaller than the BF16 source checkpoint (about a 63% reduction in disk size).
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model is ready for commercial or non-commercial use under the stated NVIDIA Open Model License and Qwen Community License 1.0. [2]
- The model checkpoint is available on Hugging Face. [3]
The Nemotron-Pre-Training-Datasets collection on Hugging Face provides large-scale data corpora used to train the NVIDIA Nemotron model family <br>
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model accepts text, image, and video input and produces text output. [3]
**Input Type(s):** Text, Image, Video <br>
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The approximately 96 GiB checkpoint is best suited for local AI hardware. [3]
The checkpoint occupies approximately **96 GiB on disk** and is best suited for local AI hardware.
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model has been validated with vLLM on NVIDIA DGX Spark and NVIDIA RTX PRO 6000 Blackwell. [3]
**Test Hardware:** NVIDIA DGX Spark and NVIDIA RTX PRO 6000 Blackwell
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The checkpoint requires one of the provided containers because upstream vLLM support for its NVFP4 scale-factor format has not yet landed. [3]
**IMPORTANT**: the NVFP4 scale factors in this checkpoint are in a new lossless compression format, but upstream vLLM support for this format has not yet landed.
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model is ready for commercial or non-commercial use. [3]
This model is ready for commercial or non-commercial use.
NVIDIA. Quoted text is outside CC BY 4.0 and keeps its original rights.
Sources
3
1
Hugging FaceOwner-controlled
https://huggingface.co/Qwen/Qwen3.8-Flash-Next
Retrieved 2026-10-27 01:17 UTC
Verified
2
NVIDIAOwner-controlled
https://huggingface.co/nvidia/Qwen3.8-Flash-Next-NVFP4
Retrieved 2026-10-27 01:17 UTC
Verified
3
NVIDIAOwner-controlled
https://huggingface.co/nvidia/Qwen3.8-Flash-Next-FP4-Local
Retrieved 2026-10-27 01:17 UTC
Verified
Versions
58| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 58 | 2026-10-27 01:17 UTC | material change | modality, availability, deployment | Current |
| 57 | 2026-10-27 01:17 UTC | material change | modality, availability, deployment | Current |
| 56 | 2026-10-26 00:38 UTC | evidence update | score | Current |
| 55 | 2026-10-26 00:38 UTC | material change | modality, availability, deployment | Current |
| 54 | 2026-10-26 00:38 UTC | material change | modality, availability, deployment | Current |
| 53 | 2026-10-25 00:47 UTC | material change | modality, availability, deployment | Current |
| 52 | 2026-10-25 00:47 UTC | material change | modality, availability, deployment | Current |
| 51 | 2026-10-24 01:17 UTC | evidence update | score | Current |
| 50 | 2026-10-24 01:17 UTC | material change | modality, availability, deployment | Current |
| 49 | 2026-10-24 01:17 UTC | material change | modality, availability, deployment | Current |
| 48 | 2026-10-23 00:58 UTC | material change | modality, availability, deployment | Current |
| 47 | 2026-10-23 00:58 UTC | material change | modality, availability, deployment | Current |
| 46 | 2026-10-22 00:28 UTC | evidence update | score | Current |
| 45 | 2026-10-22 00:28 UTC | material change | modality, availability, deployment | Current |
| 44 | 2026-10-22 00:28 UTC | material change | modality, availability, deployment | Current |
| 43 | 2026-10-20 16:32 UTC | material change | modality, availability, deployment | Current |
| 42 | 2026-10-20 16:32 UTC | material change | modality, availability, deployment | Current |
| 41 | 2026-10-19 17:30 UTC | evidence update | score | Current |
| 40 | 2026-10-19 17:30 UTC | material change | modality, availability, deployment | Current |
| 39 | 2026-10-19 17:30 UTC | material change | modality, availability, deployment | Current |
| 38 | 2026-10-18 18:19 UTC | material change | modality, availability, deployment | Current |
| 37 | 2026-10-18 18:19 UTC | material change | modality, availability, deployment | Current |
| 36 | 2026-10-17 19:10 UTC | evidence update | score | Current |
| 35 | 2026-10-17 19:10 UTC | material change | modality, availability, deployment | Current |
| 34 | 2026-10-17 19:10 UTC | material change | modality, availability, deployment | Current |
| 33 | 2026-10-16 20:05 UTC | material change | modality, availability, deployment | Current |
| 32 | 2026-10-16 20:05 UTC | material change | modality, availability, deployment | Current |
| 31 | 2026-10-15 22:33 UTC | evidence update | score | Current |
| 30 | 2026-10-15 20:56 UTC | material change | modality, availability, deployment | Current |
| 29 | 2026-10-14 21:47 UTC | evidence update | score | Current |
| 28 | 2026-10-14 21:47 UTC | material change | modality, availability, deployment | Current |
| 27 | 2026-10-13 22:39 UTC | material change | modality, availability, deployment | Current |
| 26 | 2026-10-12 23:32 UTC | evidence update | score | Current |
| 25 | 2026-10-12 23:32 UTC | material change | modality, availability, deployment | Current |
| 24 | 2026-10-12 00:28 UTC | material change | modality, availability, deployment | Current |
| 23 | 2026-10-11 01:22 UTC | material change | modality, availability, deployment | Current |
| 22 | 2026-10-10 02:15 UTC | material change | modality, availability, deployment | Current |
| 21 | 2026-10-09 03:11 UTC | evidence update | score | Current |
| 20 | 2026-10-09 03:11 UTC | material change | modality, availability, deployment | Current |
| 19 | 2026-10-08 04:06 UTC | material change | modality, availability, deployment | Current |
| 18 | 2026-10-07 05:02 UTC | material change | modality, availability, deployment | Current |
| 17 | 2026-10-06 05:53 UTC | material change | modality, availability, deployment | Current |
| 16 | 2026-10-05 06:49 UTC | material change | modality, availability, deployment | Current |
| 15 | 2026-10-04 07:45 UTC | material change | modality, availability, deployment | Current |
| 14 | 2026-10-03 08:42 UTC | material change | modality, availability, deployment | Current |
| 13 | 2026-10-02 09:37 UTC | material change | modality, availability, deployment | Current |
| 12 | 2026-10-01 10:28 UTC | evidence update | score | Current |
| 11 | 2026-10-01 10:28 UTC | material change | modality, availability, deployment | Current |
| 10 | 2026-09-30 11:22 UTC | material change | modality, availability, deployment | Current |
| 9 | 2026-09-29 12:16 UTC | material change | modality, availability, deployment | Current |
| 8 | 2026-09-28 13:05 UTC | material change | modality, availability, deployment | Current |
| 7 | 2026-09-27 14:01 UTC | evidence update | score | Current |
| 6 | 2026-09-27 14:01 UTC | material change | modality, availability, deployment | Current |
| 5 | 2026-09-26 14:55 UTC | material change | modality, availability, deployment | Current |
| 4 | 2026-09-25 15:48 UTC | material change | modality, availability, deployment | Current |
| 3 | 2026-09-24 16:39 UTC | material change | modality, access, deployment, pricing, availability, deployment | Current |
| 2 | 2026-09-24 16:39 UTC | material change | modality, access, deployment, pricing, availability, deployment | Current |
| 1 | 2026-09-23 17:33 UTC | publication | None | Current |
In briefs
1- New model releases span quantized language models and image generation Daily brief 2026-09-23
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.
