This release focuses on giving you more control during generation and making reasoning behavior clearer while streaming and after messages load.
Streaming controls you can trust
Stop now responds instantly in the UI when you click it, so you can quickly halt a response mid-stream.
Stopping also triggers server-side stream cancellation, which helps ensure the underlying request is actually stopped, not just hidden in the interface.
Reasoning display and streaming fixes
Reasoning is now shown above the main text content, making it easier to follow the model’s thought process before reading the final answer.
When reasoning is requested but the provider does not return it, the UI can now show a redacted state so you can tell the difference between “no reasoning” and “reasoning not available.”
Streaming and message reload behavior now includes reasoning and thinking duration information where available, so the reasoning panel stays consistent as content arrives.