OrcaRouter Free is a text-only endpoint in OrcaRouter's auto-routed lineup, served at an OpenAI-compatible /v1 path so it plugs into existing developer tooling. It is positioned as a no-cost entry point: input and output are both free per million tokens, with no separate cache-read tier published. The endpoint exposes a 66K-token context window and a 33K-token maximum output, giving enough room for medium-sized prompts and long generations without paying per token. Capability flags advertised on the listing cover reasoning, tool calling, structured output, attachments handling, and temperature control, suggesting the router is meant to cover everyday chat, agentic, and JSON-formatted workflows in a single endpoint.

The endpoint was released and last updated on the same day in late August 2026, and is delivered as a single non-deprecated route, making it a stable target for prototyping. Because it sits in the auto family with no disclosed vendor or open weights, it should be read as a routing abstraction rather than a fixed model: behavior, latency, and quality are likely to vary depending on which upstream model the router selects for each request. Practical fit is therefore strongest for cost-sensitive experimentation, lightweight assistants, and early-stage agent demos where zero pricing matters more than guaranteed response characteristics, while teams needing a specific model's reasoning style should pick a named upstream endpoint instead.