Skip to main content

Choosing Kaiya's Effort Level

Kaiya gives you a direct say in how hard it thinks about a task. Use the effort selector next to the message box to pick an effort level before you send a message.

For most work you can leave Kaiya on Auto. Change the effort when a brief is ambiguous, the research is more involved, or the quality bar is especially high.

The options​

Kaiya chooses between Balanced and Thinking for you, based on what you are asking.

Use Auto for most day-to-day work - it is a good default and will scale up on its own for more demanding requests.

Legacy​

The Kaiya model used before the V3 release (March 2026).

Use Legacy if a task is behaving oddly on the newer model and you want a fallback to previous behaviour.

Balanced​

Good for everyday work:

  • Running a candidate or company search
  • Drafting emails and short pieces of content
  • Following up on a conversation with refinements
  • Quick fact-checks and translations

Thinking​

Use Thinking when a task is more demanding:

  • A brief is ambiguous and needs careful interpretation
  • Research needs to be cross-checked across several sources
  • Multiple related steps depend on reasoning
  • The output is reputation-impacting (client deliverables, shortlists)

Deep Thinking​

Use Deep Thinking when quality matters far more than speed and you want Kaiya to go deep:

  • Detailed market maps or industry research
  • Sensitive or contentious topics that need nuance
  • A long list of items to evaluate consistently (more than about 10)
  • A task where you want Kaiya to double-check itself before finishing

When to move up from Auto​

Move from Auto or Balanced into Thinking or Deep Thinking when any of these are true:

  • The task is ambiguous or requires careful interpretation
  • The research crosses several sources
  • The deliverable will be seen by a client
  • You are asking Kaiya to apply the same judgement across many items
  • You are working on a high-stakes decision

If in doubt, ask Kaiya which effort level it suggests:

Given the task I'm about to give you, what effort level do you recommend?

When to fall back to Legacy​

Move to Legacy only if you notice that the newer model is handling a specific type of task less well than you remember. It is intended as a safety valve, not a default.