Pick up to three routes. Every value is read from the same registry that prices your requests — pricing version 2026-08-24.1. Unverified fields stay unverified: comparison never fills gaps with estimates.
Comparing 2 of 3 possible routes.
| Attribute | auto | auto:cheap |
|---|---|---|
| Type | Intent — Fleet selects | Intent — Fleet selects |
| Billing tier | frontier | fast |
| Context window | 1M tokens | 200K tokens |
| Max output | 128K tokens | 64K tokens |
| Input / MTok | $6.25 | $1.25 |
| Output / MTok | $31.25 | $6.25 |
| Caller tools | Relayed | Relayed |
| Hosted tools | None — hosted_tools: [] | None — hosted_tools: [] |
| Streaming | Rejected with 400, not buffered | Rejected with 400, not buffered |
| Capabilities | reasoning, coding, long context, structured output | fast, cheap, structured output |
| Latency | Not measured | Not measured |
| Live availability | Not probed | Not probed |
Unverified fields stay unverified here. Latency and live availability are shown as not measured because this deployment performs no latency sampling and no health probe — comparison never fills a gap with an estimate.
Tier is what you are billed at, not a quality ranking. An intent reserves at the most expensive tier it may resolve to and settles at the tier of the model that actually answered, so a frontier-ceiling intent frequently costs less than its reserve. See pricing.
No benchmark scores, no latency percentiles, no uptime figures. Modavio does not measure them, and publishing a number nobody sampled would make this table less useful than leaving the row honest.