LM Studio added LM Link to its Locally iPhone app, so your phone can now talk to language models already running on your Mac.
LM Studio, a widely used desktop tool for running large language models locally, updated its Locally companion app with a feature called LM Link. The feature bridges an iPhone to whatever LLMs are loaded and running on a Mac, so the phone handles the interface while the Mac does the actual inference. None of the conversation leaves the local network — no API calls, no cloud routing, no usage logs sitting on a third-party server. The update shipped earlier this week.
The obvious alternative — running LLMs directly on an iPhone — is improving but still constrained by mobile hardware. Apple and others have pushed smaller on-device models forward, but anything above a few billion parameters demands sustained compute and memory that phones weren't built to maintain. LM Link sidesteps that ceiling entirely by treating the Mac as a local inference server and the phone as a thin client, which is a cleaner division of labor than waiting for mobile silicon to catch up.
Whether most people genuinely need to query a local Mistral model from the couch rather than the desk is worth asking — but the privacy argument is harder to wave away.