Routing methodology
How Oran T1 picks a model.
T1 reads each prompt and routes it to the model best-suited to answer it. The decision happens before the message reaches any frontier model. Every reply you receive carries an attribution chip showing which model T1 picked.
What goes into the decision
- Content type. Reasoning, writing, code, analysis, vision, summary.
- Complexity. Short factual question vs. multi-step reasoning.
- Context length. Conversations and attached documents.
- Modality. Whether an image is attached.
- Tier. Which models are available on your plan.
Where the landing page numbers come from
The capability matrix on the home page shows relative strengths across five categories — reasoning, writing, code, long context, and speed/cost. The bars are illustrative, not precise scores. They reflect typical relative ordering from public evaluations including HumanEval, MMLU, MT-Bench, and long-context retrieval benchmarks. We'll expand this page with specific sources and our own evaluation harness as it lands.
Questions about routing? hello@oran.chat