🚀 New Features
- Download decision models from the Hub and run them directly on the TurboQuant page for local evaluation of decision workflows.
- ChatGPT models now expose a reasoning-effort indicator so you can see how much "thinking" each response used.
🔧 Improvements & Fixes
- Downloads: queued and cancelled download states are quieter to reduce UI noise.
- Model browser and picker: compacted and unified UI with shorter technical names, real lead-model summaries, device-fit badges, preserved quant selection, outlined active/unload icons, and tighter selector feedback and motion for clearer model management.
- Sidebar/layout: Images is now a full-row section toggle and collapsed side panels are recovered to preserve your layout.
- Connectors: connector status is shown in the byline and Exa fallback is reported as usable so connector health is clearer.
- Cloud: unsupported sampling controls are hidden to prevent invalid configurations.
- Updater: release notes get more room, previews behave better (including build-only and local previews), and the updater stacks above active downloads for a cleaner update flow.
- Reasoning & agents: default required thinking is lowered and tool activity details are tightened for clearer agent behavior.
- 10+ small stability and UI fixes across downloads, models, layout, and the updater.
🙏 Contributors
Thanks to @Vect0rM for their contributions to this release!