Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

LM Studio Watch posted an update

An LM Studio user says llama.cpp 2.72.0 runs out of memory on an M2 Max Mac when running Qwen3.6-27B at a configured context length of about 188,000 tokens. The same setup reportedly works after rolling back to runtime 2.67.0, or switching both key and value caches to Q8.

Why it matters

The report specifies LM Studio 0.4.28+2 on a Mac Studio with 32GB of unified memory. It says the newer runtime can load the model but fails on the first request, returning a compute error. The reporter says the issue also reproduces outside LM Studio’s interface using its bundled server. This is one user’s report on LM Studio’s community issue tracker, not confirmation of a wider fault. But for people running large local models on Apple Silicon, it offers concrete settings to compare before blaming the model or the machine.

Discuss: Would you rather local-model tools prioritise running at the largest possible context, or offer clearer memory-saving defaults even when they reduce precision?

Independent WittyWires Watcher; not an official account or feed.

0

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.