The deepest, slowest, most thoughtful part of Arka is now one tap away.
Until this release, Adaptive Thinking lived in a brain-shaped button next to the textarea. It was discoverable, but it was also a permanent fixture — visible even when you didn't need it. Cluttered the input bar. We moved it.
Where Adaptive Thinking is now
Tap the model selector inside the chat box — the chip that says “Arka Helix 1.” The dropdown shows two things: the current model, and a toggle labelled Adaptive thinking, with a short description (“Thinks longer on complex tasks”). Flip it on, send your message, and Arka switches into deep-reasoning mode for that response.
The state persists across the conversation. If you flip it on once, every message after that uses adaptive thinking until you flip it back off — or hit your daily limit on Free.
When deep reasoning is worth it
Adaptive thinking is slower and more expensive (in tokens, and in your daily limit). It is worth it for:
- Multi-step math or logic problems
- Code that needs to reason about edge cases
- Analysis questions with several constraints to balance
- Anything where the wrong answer would cost you serious time to undo
For “what does this acronym mean” or “rewrite this email,” leave it off. Helix 1 is plenty. For a deeper dive into how Adaptive Thinking changes the output, see Deep Reasoning, explained.