4 ms·

LM Link: Use local models on remote devices, powered by Tailscale

• SeriousM 8mo ago
  Wow, that's huge! I'll try that at home.
• ZeroCool2u 8mo ago
  I've been trying to use this all morning, but I keep getting 500/auth errors, even on a completely new device. Can't even login to my LM Studio account right now.
• 7777777phil 7mo ago
  The main thing i see here is that prompts never touch a third-party server. If you're in a regulated industry or just don't want proprietary context hitting an API, running inference on your own hardware with encrypted p2p from any device is really cool (and useful.) (staying in userspace via tsnet without touching kernel sockets is a nice touch too.)
• solarkraft 7mo ago
  Cool idea, but I keep wondering why there’s no browser version of LM Studio. LM Studio is cool, but setting everything up on ONE device is of no value to me.