Model prices across LLM gateways
Choose a model and compare the estimated cost of the same token workload through each gateway.
Catalog prices checked Oct 18, 2026View all models
Choose a model
Prices use one fixed workload so the gateway totals are directly comparable.
Model
Same model · same workloadGPT-6 AstraOfficial API baseline$60Standard · up to 272K input per request
OpenRouter1M input + 1M output$63.3Model usage$60Credit purchase fee estimate+$3.35.5% when purchasing credits, not per request
Vercel AI GatewayLowest gateway cost1M input + 1M output$60Model usage$60Gateway markup$0No token markup
Cloudflare AI Gateway1M input + 1M output$63Model usage$60Unified Billing fee estimate+$35% unified billing fee
Portkey1M input + 1M output$60 + planModel usage$60Platform planSeparate planProvider usage plus a separate platform planComparison tier: Standard · up to 272K input per request
| Price conditions | Input / 1M | Cached input / 1M | Output / 1M |
|---|---|---|---|
| Standard · up to 272K input per requestUsed above | $10 | $1 | $50 |
| Standard · over 272K input per request | $20 | $2 | $75 |
The applicable rate depends on the request region, billing time, and input length. The total above still uses the shared comparison basket of 1M input and 1M output tokens.
How the comparison works
Gateway catalogs publish input and output rates per 1 million tokens. The comparison basket uses 1 million input tokens plus 1 million output tokens so each rate appears once in the calculation. It is a neutral benchmark, not a claim about typical request size. We then add the gateway's published platform fee; taxes, payment charges, discounts, and monthly plans remain separate.
Gateway platform details