Sniffed the network traffic of Ollama and LM Studio: chats stay local, apps don't
TheOyinbooke · x · 2026-10-14
The author monitored every outside connection Ollama, LM Studio and Bionic opened on a Mac Studio. Result: LM Studio opens three connections to its own servers at launch and Ollama reaches ollama.com within a second, but during chats neither app makes a single new outside connection — your prompts truly stay local. "Local" is a promise about prompts only: LM Studio's model searches go out as anonymized queries and its update check sends app version, build, OS and IP; Ollama pulls models over HTTPS and auto-updates. Neither documents a switch to disable update checks.
More from Infra
- Prefix Sliding paper: training-free 3x speedup for long reasoning via prefix + sliding window — cephaloform · 2026-10-14
- Mesa merges YOLOX with PyTorch support, expanding NPU capabilities — Fcking_Chuck · 2026-10-14
- Dev complains HDDs grind his server to a halt as NVMe markup hits 67,000% — haydendevs · 2026-10-14
- First real-world look at Google's Ironwood TPU rack at Google Cloud event — firstadopter · 2026-10-14
- Samsung shows off HBM5 samples and Custom HBM with xPU stacked design — firstadopter · 2026-10-14
- AWS launches Lambda Web Functions with up to 128 concurrent requests per environment — tlakomy · 2026-10-14