Downloads · 30 days
25
8% of all-time downloads
Qwen3-30B-A3B-FP8-OC
hyper-accel/Qwen3-30B-A3B-FP8-OC
Qwen3-30B-A3B-FP8-OC is a machine learning model from hyper-accel. 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.
Derived from Qwen/Qwen3-30B-A3B at revision ad44e777bcd18fa416d9da3bd8f70d33ebb85d39.
All-time downloads
315
Public
Parameters
30.7B
31.4 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors31.3 GB · 100%
Parameter types
How the weights are stored.
I830.1B · 98%
Base models
Model card
From the Hugging Face model README
Qwen3-30B-A3B FP8-OC
Derived from Qwen/Qwen3-30B-A3B at revision ad44e777bcd18fa416d9da3bd8f70d33ebb85d39.
LLM Compressor quantizes attention and expert matrix weights to E4M3FN with one FP32 scale per output channel. HART stores those FP8 bytes and scales in the Bertha EVT0 packed weight format. Other model parameters remain BF16. This checkpoint requires the HyperAccel vLLM plugin.
- Type
- qwen3_moe
- License
- apache-2.0
- Created
- Aug 24, 2026
- Updated
- Aug 25, 2026