OpenAI
OpenAI's Responses, Chat Completions, embeddings and moderation endpoints, metered per token against the same balance as the rest of the catalog, with gpt-5.6's cache- and context-tiered Standard rates read straight from each response.
Where OpenAI lands
Composite scores out of 10, from benchmark runs where every provider faces the identical task corpus. Rank is within that capability's field. Full method: how we measure.
OpenAI sits outside the web-access benchmark — the search/scrape/crawl corpora don't exercise a model inference API, so this page carries catalog facts without a scorecard.
The honest pitch
Pick it when the task calls for an OpenAI model by name and per-token spend should land on the one ledger you already watch.
Skip it for anything outside the gateway's admitted surface: only a fixed, maintained model list and a closed set of request fields are reachable, and everything else is rejected before it reaches OpenAI.
About OpenAI
OpenAI’s inference surface sits behind a policed route rather than a plain forward: the model has to be on a maintained catalog before a request is even admitted, a closed set of request fields is allowed through, and everything else gets rejected before it reaches OpenAI. Files, vector stores, batches and fine-tuning jobs are tracked per account, so one customer’s stored data or running job can never be read or touched through another customer’s key. Usage is billed from the token counts each response reports, and an account can register its own OpenAI key to bill that provider directly instead of metering through the gateway.
The current flagship line is gpt-5.6, shipped in three sizes — Sol, Terra and Luna — with the bare gpt-5.6 model id an alias for Sol, per OpenAI’s model page. Standard-tier rates split on context length at 272,000 tokens, and a cache write is its own billing bracket rather than an add-on to the input charge: each token lands in exactly one of input, cached input, or cache write, never more than one.
| Model | Context | Input | Cached input | Cache write | Output |
|---|---|---|---|---|---|
Sol (gpt-5.6) |
≤272K tokens | $5.00 | $0.50 | $6.25 | $30.00 |
Sol (gpt-5.6) |
>272K tokens | $10.00 | $1.00 | $12.50 | $45.00 |
| Terra | ≤272K tokens | $2.00 | $0.20 | $2.50 | $12.00 |
| Terra | >272K tokens | $4.00 | $0.40 | $5.00 | $18.00 |
| Luna | ≤272K tokens | $0.20 | $0.02 | $0.25 | $1.20 |
| Luna | >272K tokens | $0.40 | $0.04 | $0.50 | $1.80 |
Rates are per 1M tokens at OpenAI’s Standard tier; Batch runs every one of those figures at half price. Older families — gpt-5.5, gpt-4.1, the o-series and the rest — stay reachable through the same maintained catalog at their own published rates. See OpenAI’s pricing and prompt caching docs for the full model list and cache-write mechanics.
Reaching it through NativePort
Identical paths, parameters and responses to OpenAI's own documentation — the gateway holds the upstream credential and meters your balance at OpenAI's real published usage price.
curl https://api.nativeport.ai/openai/<native-path> \ -H "Authorization: Bearer $NATIVEPORT_API_KEY"