Conservative reading frame
The cheapest model is not always the best model. The right low-cost choice depends on the shape and risk level of the workload.
Compare with the same token assumptions
Do not treat unknown prices as zero
Verify the official source before production
Avoid headline-price traps
A route with very low input price can still be expensive if output is high, context is too small, or production limits make it unreliable.
Treat free routes carefully
Free routes are useful for experimentation, but production decisions should check limits, uptime, terms, and whether the route can disappear or change.
Filter by required capability
Do not compare text-only models against workloads that require vision, JSON, tool calling, audio, or long context.
