OpenAI rolls out GPT‑5.5

OpenAI said on April 23 that GPT-5.5 is rolling out across paid ChatGPT plans, with GPT-5.5 Pro reserved for higher-tier users who need more compute. (openai.com) The company said GPT-5.5 is available in ChatGPT and Codex for Plus, Pro, Business and Enterprise users. GPT-5.5 Pro is rolling out in ChatGPT to Pro, Business and Enterprise customers. (openai.com) A day later, on April 24, OpenAI updated the release to say GPT-5.5 and GPT-5.5 Pro were also available in the application programming interface, or API. The API changelog says GPT-5.5 Pro is offered through the Responses API, including Batch, for harder tasks. (openai.com) (developers.openai.com) Large language models predict and generate text, code and other outputs from patterns in data; newer versions are usually judged on how well they reason across long prompts, use tools and finish multistep jobs. OpenAI is pitching GPT-5.5 as a model for “real work” on a computer, including coding, web research, spreadsheets and software use. (openai.com) OpenAI said GPT-5.5 keeps GPT-5.4’s per-token latency while improving performance on coding, browsing, math and computer-use tests. In OpenAI’s release table, GPT-5.5 scored 82.7% on Terminal-Bench 2.0 versus 75.1% for GPT-5.4, and 84.4% on BrowseComp versus 82.7%. (openai.com) The company said the Pro version uses the same underlying model with a setting that applies parallel test-time compute, meaning it spends more compute checking paths before answering. OpenAI’s system card says that can change risk in some cases, so GPT-5.5 Pro was separately evaluated for some safety areas. (openai.com) OpenAI also said GPT-5.5 supports a 1 million-token context window in the API, along with image input, function calling, prompt caching, web search and built-in computer use. The API changelog says GPT-5.5 defaults to medium reasoning effort. (developers.openai.com) In Codex, OpenAI said GPT-5.5 is the recommended model for implementation, refactors, debugging, testing and validation. The Codex changelog says users may need to update the command-line tool, integrated development environment extension or app before the model appears. (developers.openai.com) The launch came eight days after OpenAI’s April 16 release post for GPT-5.4, which it had described as its most capable and efficient frontier model for professional work. That tighter release cadence suggests OpenAI is shipping model updates faster across ChatGPT, Codex and the API than it did earlier in the GPT-4 era. (openai.com 1) (openai.com 2) OpenAI said it tested GPT-5.5 with internal and external red teamers, added checks for advanced cybersecurity and biology capabilities, and gathered feedback from nearly 200 early-access partners before release. The company is framing the model less as a chatbot upgrade than as software that can take a messy task and keep working through tools until it is done. (openai.com)