Tools
Ollama v0.34.0-rc1 lets ChatGPT Desktop run local Ollama models
Ollama cut v0.34.0-rc1 on 2026-09-05 at 23:49 UTC, and the headline item is that Ollama-hosted open models can now be selected directly inside ChatGPT Desktop, with setup driven from the Ollama macOS app. The same release improves structured-output performance on Apple Silicon and adds OpenAI-compatible client tool search plus response compaction, fixing images that previously broke through compacted responses. For builders this is the first time OpenAI's own desktop client is a front-end for a local inference server, which changes the calculus on running open weights without giving up the UI people already use.
Source
↳ Follow the thread