GPT-5.5 Pro — Specs, Pricing & Limits

GPT-5.5 Pro's price is the second thing worth noticing about it. The first is its entry-tier request ceiling, which sits an order of magnitude below what the standard models get at the same tier — a limit it now shares with GPT-5.4 Pro, whose rates OpenAI cut to match after August.

That ceiling is easy to miss when you compare models on price, since nothing about a high per-token rate signals a correspondingly tight request rate. For a workload that calls this model frequently rather than occasionally, the Build row is the number to read first: at the lower usage tiers it is the thing most likely to constrain a real deployment. OpenAI also warns that Pro requests can take several minutes and suggests background mode to avoid timeouts.

Its reasoning effort is narrower than any standard model's — medium, high and xhigh, with high as the default — and there is no Fast mode and no cached-input price. The long-context repricing rule is not stated on this model's own page; the pricing page does publish a long-context Standard band for it, but not the full-request wording, so this site doesn't model a cliff for it.

If what you want is Pro-style extra model work rather than this specific model, the GPT-5.6 and GPT-6 families offer reasoning.mode: "pro" on the Responses API, billed at the selected model's own standard rates — more tokens, not a higher rate.

The table above is computed directly from this site's facts module: pricing across every published service tier and the rate-limit table.