AI Orchestrator
DE Early Access

Decision model

Model Routing

Choose the tool, model and reasoning mode deliberately — not the most expensive option by default.

Reasoning mode names and availability vary by provider and change over time. The names below are a configurable decision model, not a claim about any specific vendor’s current offering.

Task categories

Route by the kind of task.

  • Strategy and product decisions

    Direction, prioritization and scope. Favors strong reasoning and broad context over raw speed.

  • Implementation

    Building and changing code or content within clear scope. A capable model at a justified reasoning level.

  • QA and verification

    Checking that a change does what it should. Often a focused, scoped pass rather than the largest mode.

  • Research

    Gathering and synthesizing information before deciding. Breadth matters more than maximum depth.

  • Critical multi-agent review

    High-stakes changes verified by independent passes. The case where extra reasoning and redundancy are justified.

Reasoning modes

Escalate only when the task earns it.

  1. Normal Everyday tasks with clear scope and low risk.
  2. High Harder reasoning, ambiguity, or wider context.
  3. Extra Deeper analysis when correctness is sensitive.
  4. Max The hardest single-pass problems.
  5. Ultracode Multi-agent orchestration for the most demanding work.
Do not use the most expensive mode for every task. Use the smallest mode that can safely solve the problem.