Downloads · 30 days
38
30% of all-time downloads
Qwen3.6-35B-A3B-FP8-Dynamic
huseinzolkepliscicom/Qwen3.6-35B-A3B-FP8-Dynamic
Qwen3.6-35B-A3B-FP8-Dynamic is a machine learning model from huseinzolkepliscicom. 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.
FP8-dynamic (compressed-tensors) quantization of Qwen/Qwen3.6-35B-A3B (MoE + gated-DeltaNet linear attention + multimodal), via llm-compressor. The MoE expert + full-attention Linear layers are FP8; kept in bf16: the…
All-time downloads
127
Public
Parameters
35.1B
73.5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors37.7 GB · 100%
Parameter types
How the weights are stored.
F8_E4M332.6B · 93%
Base models
Model card
From the Hugging Face model README
Qwen3.6-35B-A3B — FP8 Dynamic
FP8-dynamic (compressed-tensors) quantization of Qwen/Qwen3.6-35B-A3B (MoE + gated-DeltaNet
linear attention + multimodal), via llm-compressor.
The MoE expert + full-attention Linear layers are FP8; kept in bf16: the vision tower, the MoE
routers and shared-expert gates, and the GDN linear-attention layers (quantizing any of these
breaks vLLM loading or output quality). Verified serving on vLLM 0.23.0:
vllm serve huseinzolkepliscicom/Qwen3.6-35B-A3B-FP8-Dynamic \
--tensor-parallel-size 1 --gpu-memory-utilization 0.9
- Type
- qwen3_5_moe
- License
- apache-2.0
- Created
- Jul 15, 2026
- Updated
- Jul 15, 2026