Instructions to use unsloth/Qwen3.8-27B-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Local Apps Settings
- Unsloth Desktop
Updated files for UD3.0 final?
Assume these will be landing soon? Fingers crossed :)
Edit: I think perhaps not a thing as the quant isn't dynamic for these and the 3.0 final for UD was just a quant rebalance
When UD3.0 nvfp4?
Might be no change for nvfp4
I think perhaps not a thing as the quant isn't dynamic for these and the 3.0 final for UD was just a quant rebalance
As far as I understand, Unsloth Dynamic quants is their technique for GGUF format. And I'd assume it doesn't apply to this NVFP4 quant in safetensors format.
Also (as per https://unsloth.ai/docs/basics/dynamic-3.0-ggufs#not-overfitting ), I think UD 3.0 is now focused on smaller quants, as they mentioned that they still use UD 2.0 for larger quants.
They don't mention what is considered small or large, but my intuition tells me smaller is 1-3bit quants, larger is 4-8bit quants.
In any case, I'm sure they'll keep us posted on any improvements in their larger UD quants.
As far as I understand, Unsloth Dynamic quants is their technique for GGUF format. And I'd assume it doesn't apply to this NVFP4 quant in safetensors format.
Also (as per https://unsloth.ai/docs/basics/dynamic-3.0-ggufs#not-overfitting ), I think UD 3.0 is now focused on smaller quants, as they mentioned that they still use UD 2.0 for larger quants.
They don't mention what is considered small or large, but my intuition tells me smaller is 1-3bit quants, larger is 4-8bit quants.In any case, I'm sure they'll keep us posted on any improvements in their larger UD quants.