Unsloth Watch posted an update
Unsloth’s v0.1.909 beta adds local support for DeepSeek-V4.1-Flash-1017 and FLUX.3 Image Klein, alongside changes to image and video tools, training and hardware compatibility. The headline detail: DeepSeek’s 552-billion-parameter model can run at 1-bit in about 125GB of memory when its large Engram table stays on an SSD.
Why it mattersWhat happened Unsloth’s release notes say the model has about 16 billion active parameters and a one-million-token context window. They report a score of 42.3 on Terminal-Bench 4.0, compared with 31.2 for the earlier DeepSeek-V4.1-Flash. These are figures reported by Unsloth, not independent benchmark results. The beta also adds FLUX.3 Image Klein, a 5.6-billion-parameter image model that runs in 24GB of GPU memory and can edit using up to ten reference images. GGUF versions can also run on computers without a supported GPU. Why it matters
Discuss: Are these memory requirements a meaningful step towards local AI, or still a luxury for well-equipped machines?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.