Quality tiers

T1 / T2 / T3 - cost vs quality tradeoffs, per modality.

Every multimodal request is associated with a quality tier (T1, T2, or T3). The tier controls which provider and model are used when quality_mode is auto. In strict mode the tier is informational only - you pick the exact model yourself.

Tier cost references

These are reference prices for the tier routing logic. Your actual cost depends on the provider's published rate at the time of the call.

Image - T1 ~$0.003/img · T2 ~$0.025/img · T3 ~$0.05/img
Video - T1 ~$0.006/s · T2 ~$0.04/s · T3 ~$0.12/s
Audio TTS - T1 ~$15/1M chars · T2 ~$22/1M chars · T3 ~$99/1M chars
Embeddings - T1 ~$0.02/1M tokens · T2 ~$0.06/1M tokens · T3 ~$0.13/1M tokens

Quality modes

strict (default) - you specify provider + model on every call. HiWay never changes anything.
guided - you still control the model, but HiWay surfaces tier-optimization suggestions in Analytics → Guided.
auto - HiWay routes within the configured tier. For video T2/T3, a T1 preview is generated first for your approval.

HiWay never silently downgrades

If the requested tier is unavailable (e.g., no matching key for the T2 provider), the API returns a 422 with a clear error - it does not fall back to T1 without your consent.