Question 1

How does RightOne choose a model for my prompt?

Accepted Answer

AutoRouter classifies the subject, depth, reasoning needs, and difficulty. It ranks eligible models using measured performance and current pricing. Server-side account and affordability policy then selects one model for the turn.

Question 2

Is it actually cheaper than just using one big model?

Accepted Answer

In our disclosed June 2026 benchmark, RightOne routed 39 of 50 tasks to free models and cost 64% to 81% less than three flagship-only baselines. That result belongs to this task mix. It changes with prompts, available models, provider pricing, and routing updates.

Question 3

Does routing make the answers worse?

Accepted Answer

The same benchmark scored RightOne 9.28 out of 10 overall, compared with 9.64 to 9.68 for the flagship-only baselines. Math, physics, and code were within 0.1 points. Full-stack work and open conversation showed larger gaps, so we do not claim universal quality parity.

Question 4

Does the routing step add latency?

Accepted Answer

Yes, classification adds work before the model call. We have not published a benchmark showing a universal latency reduction. End-to-end speed depends on the selected model, provider load, network conditions, and response length.

Route to the Right One. RightOne ChatOS