OpenAI released GPT-5.5 on April 23, describing it as its most capable model to date and the first of what the company calls a new class of intelligence. The model is rolling out in ChatGPT and in Codex, OpenAI’s coding agent, to Plus, Pro, Business and Enterprise subscribers, with API access following.
The launch comes roughly seven weeks after GPT-5.4, an unusually short interval even by the current pace of the industry, and it arrived after several days of speculation fueled by leaked references in Codex, a brief API routing glitch that exposed an unreleased model, and a cryptic social media post from company executives.
What OpenAI claims
In its announcement, OpenAI said GPT-5.5, developed under the internal codename Spud, delivers its largest gains in agentic coding, knowledge work and scientific research. Agentic coding refers to a model working through multi-step programming tasks with limited human direction: reading a codebase, planning changes, running tests and correcting its own mistakes.
Greg Brockman, OpenAI’s president, characterized the model as a step beyond incremental improvement, telling reporters it represents a new class of intelligence rather than a faster version of the same thing, according to The New Stack. The company published benchmark results showing improvements over GPT-5.4 on coding and reasoning tests, though as with earlier releases those figures are OpenAI’s own and have not yet been independently replicated.
Higher prices for API users
The new model costs more to use. Several outlets, including Fortune, reported that API pricing for GPT-5.5 is roughly double that of its predecessor, a sign that OpenAI is positioning the model for high-value automated work rather than casual chat. ChatGPT subscribers get access at their existing subscription prices.
Context
The release underscores how quickly frontier labs are now shipping. OpenAI, Anthropic and Google have each moved to release cadences measured in weeks, with a growing share of each update aimed at agents that complete tasks end to end rather than at conversational quality alone. Prediction markets had priced a GPT-5.5 launch in April as highly likely in the days before the announcement, reflecting how closely the industry watches OpenAI’s schedule.
Whether the gains OpenAI reports translate into reliable autonomous work in real settings will become clearer as developers put the model through independent testing in the coming weeks.
