Was this created with the fixes in mlx-lm version 0.30.8?
This model mlx-community/Qwen3.5-397B-A17B-8bit was converted to MLX format from Qwen/Qwen3.5-397B-A17B using mlx-lm version 0.30.8.
As the pull request to increment mlx-lm's version number to 0.30.8 was created today at 9:23 AM EST, and the upload of this repository started yesterday night around 10 PM, it must have been made using a version of mlx-lm from someone's private branch. Does it have the important fix for Qwen 3.5 in pull #902?
Yes, this quant was created after Awni had merged pull request 902.
Releasing a version with fixed scope, and bumping the version number for the next version are separate actions. If the version number is not "bumped" when a release happens, it happens that the version numbers of mlx-lm are not always up-to-date.
They had released 0.30.7 already, but did not yet bump up the version number to 0.30.8 for any new code. Knowing that they will have to bump the version number to 0.30.8, I already wrote 0.30.8 in the model card as this is what people should use. Version 0.30.7 that was already released did not have PR #902 included, so I would be giving wrong information if I wrote 0.30.7 in the model card.