OpenAI GPT-5.5 API 1M-token context

OpenAI’s own product pages appear to have settled one part of this week’s GPT-5.5 rumor cycle: the 1 million-token API context window is no longer just a leak. OpenAI said in an April 23 product post that GPT-5.5 would soon be available in the Responses and Chat Completions APIs with a 1 million-token context window, and an April 24 update on the same page said GPT-5.5 and GPT-5.5 Pro were available in the API. Tech Edu Byte’s May 23 write-up described GPT-5.5 as supporting a one million-token context window and up to 128,000 output tokens, with an emphasis on structured JSON reliability and enterprise workloads. But the core capacity claim now matches OpenAI’s own public documentation, which lists GPT-5.5 with a 1.05 million context length and 128,000 maximum output tokens on its API platform pages. (openai.com) ### So what, exactly, is confirmed now? OpenAI’s April 23 announcement said GPT-5.5 would be offered to API developers at $5 per 1 million input tokens and $30 per 1 million output tokens, with a 1 million context window. An April 24 update on that same announcement said the model was available in the API. (openai.com) OpenAI’s API model documentation now shows GPT-5.5 pricing, and separately notes that prompts above 272,000 input tokens are billed at higher rates for the full session. The developer docs do not describe that as a leak or preview; they present it as current product documentation. (openai.com) ### Where did the 128,000 output-token figure come from? OpenAI’s API platform page lists GPT-5.5 with “128K max output tokens.” The same page lists GPT-5.4 with the same maximum output figure, while GPT-5.5 carries higher token pricing. Third-party model trackers have echoed that configuration. OpenRouter’s GPT-5.5 benchmark page describes the model as having roughly 922,000 input tokens plus 128,000 output tokens, which together amount to a little over 1 million tokens of total context. (developers.openai.com) ### Why did this show up as a “leak” story this week? (openai.com) Tech Edu Byte’s May 23 article appears to be part of a broader wave of GPT-5.5 commentary that mixed official specs with rumor framing. A separate May 2026 roundup from Thicket said leak chatter had focused on release timing, routing architecture and performance claims, while OpenAI’s own pages had already published baseline API details including the 1 million-token context window. (openrouter.ai) OpenAI’s public materials did not foreground the structured-JSON angle in the same way as Tech Edu Byte. Instead, OpenAI described GPT-5.5 as aimed at coding and professional work, and its latest-model guide says the model is better at preserving constraints and producing concrete next steps in complex workflows. (trends.thicket.sh) ### What does OpenAI itself say GPT-5.5 is for? OpenAI’s April 23 release said GPT-5.5’s gains were strongest in agentic coding, computer use, knowledge work and early scientific research. The company also said the model matched GPT-5.4’s per-token latency in real-world serving while using fewer tokens on some coding tasks. (developers.openai.com) The developer guide says GPT-5.5 supports the API features already available with GPT-5.4, including prompt caching, hosted tools, tool search, compaction and phase handling for manually replayed assistant items. OpenAI’s pricing page and model docs position it as the flagship option for more complex professional workloads. (openai.com) ### What should developers watch next? OpenAI’s current documentation is the clearest place to watch for changes in limits, pricing and feature support. The company’s model page, pricing page and latest-model guide now all list GPT-5.5 as an API model, and those pages are where any future revisions to context limits, billing thresholds or supported features would likely appear first. (developers.openai.com 1) (developers.openai.com 2)