|
|
--- |
|
|
language: |
|
|
- en |
|
|
- zh |
|
|
license_link: https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct/blob/main/LICENSE |
|
|
library_name: transformers |
|
|
license: apache-2.0 |
|
|
pipeline_tag: text-generation |
|
|
base_model: |
|
|
- Qwen/Qwen3-Next-80B-A3B-Instruct |
|
|
--- |
|
|
# Qwen3-Next-80B-A3B-Instruct-NVFP4 |
|
|
|
|
|
**Quantized version of [Qwen/Qwen3-Next-80B-A3B-Instruct](https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct)** using **LLM Compressor** and the **NVFP4** (E2M1 + E4M3) format. |
|
|
|
|
|
**This time it actually works!** *We think* |
|
|
|
|
|
This should be the start of a new series of *hopefully optimal* NVFP4 quantizations as capable cards continue to grow out in the wild. |
|
|
|
|
|
--- |
|
|
|
|
|
## Model Summary |
|
|
|
|
|
| Property | Value | |
|
|
|-----------|--------| |
|
|
| Base model | Qwen/Qwen3-Next-80B-A3B-Instruct | |
|
|
| Quantization | NVFP4 (FP4 microscaling, block = 16, scale = E4M3) | |
|
|
| Method | Post-Training Quantization with LLM Compressor | |
|
|
| Toolchain | LLM Compressor | |
|
|
| Hardware target | NVIDIA Blackwell (Untested on RTX cards) / GB200 Tensor Cores | |
|
|
| Precision | Weights & activations = FP4 • Scales = FP8 (E4M3) | |
|
|
| Maintainer | **RESMP.DEV** | |
|
|
|
|
|
--- |
|
|
|
|
|
## Description |
|
|
|
|
|
This model is a drop-in replacement for Qwen/Qwen3-Next-80B-A3B-Instruct that runs in **NVFP4 precision** |
|
|
Accuracy remains within ≈ 1 % of the FP8 baseline on standard reasoning and coding benchmarks. |
|
|
|
