xavierbr lines up Claude Opus 5 with more control, but the move stalls
The proposal would have modernised Mike's chat model choices while making repeat conversations cheaper to run and easier to monitor.
xavierbr proposed replacing a retired Claude option with Claude Opus 5, Anthropic's latest high-capability model, and carrying existing saved choices forward. Users would also have been able to choose how much reasoning effort the model applies, with a middle setting as the default.
- Reasoning controls: Low, Medium and High options for chat, project and table-based work.
- Request caching: reuse stable instructions and tool descriptions on repeat requests, intended to reduce repeated processing.
- Cache reporting: record when that reuse occurs, giving the team a clearer view of efficiency.
- Input checks: reject unsupported reasoning settings consistently across chat experiences.
The pull request was closed without merging, so none of this is live in Mike's target branch.
So what Legal teams watching AI cost, responsiveness and answer quality should track this if xavierbr revives it, but should treat it as an abandoned proposal for now.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?