Version 3.1.6
- New Latest AI models — GPT-5.5 and the GPT-5.4/5.3 family, Claude Fable 5 and Opus 4.7/4.8, Gemini 3.5 Flash and 3.1 Pro — are now selectable, and the model list automatically stays in sync with the MxChat core plugin going forward
- Improved Retired models (Gemini 1.5/2.0, Grok 2, older Claude) are removed from the picker and any saved selection is migrated to its current equivalent automatically
- Fixed Image analysis with Grok was failing — it now uses xAI's current Grok 4 vision model
- New Image analysis now works with Claude and Gemini — choose "Claude Vision (Anthropic)" or "Gemini Vision (Google)" in the Admin Assistant's image-analysis setting. Previously only Grok and GPT-5 vision were offered, so admins whose only configured key was Claude or Gemini had no image-analysis path.
- Fixed The Admin Assistant's model picker now shows the same up-to-date model lineup as the main MxChat plugin. Previously the picker rendered an outdated, hardcoded list that omitted the newest models (GPT-5.5, Claude Fable 5 / Opus 4.8, Gemini 3.5 Flash and more) and could list retired models that errored with "Invalid model ID" when chosen.
- Fixed PHP 8.2 deprecation notices ("Creation of dynamic property") no longer appear in debug mode
- Improved The Admin Assistant model picker now shows every AI model even for providers you have not added an API key for yet, so you can see your options before deciding which provider to set up. Models that still need a key are marked, and the key is only required when you actually chat — matching how the main MxChat plugin works.
- New When the selected Assistant model belongs to a provider you have not added an API key for, a clear note now appears above the message box naming exactly which provider key to add and linking straight to MxChat Settings — so you know before sending rather than after a failed message.
