OpenAI Releases GPT-5.5 "Spud" — Agentic Coding Focus, 900M Weekly Users

OpenAI has released GPT-5.5, a new model the company is positioning around agentic coding and long-context work. The announcement, published on OpenAI’s site on April 23, 2026, says the model is meant to handle more complex software tasks, with support for a 1 million token context window and improved efficiency over GPT-5.4.

The release lands only a week after Anthropic’s Opus 4.7, which makes the timing notable for anyone tracking the fast pace of model launches. OpenAI did not frame GPT-5.5 as a general-purpose consumer refresh alone. Instead, the company emphasized use cases where the model needs to plan, edit, reason across large codebases, and keep track of long conversations or documents without losing earlier context.

That 1 million token context window is one of the headline technical details. In plain terms, context is the amount of text a model can consider at once, including prompts, files, and prior messages. A larger window matters for tasks like reviewing large repositories, working through multiple documents, or maintaining continuity across longer agent runs.

OpenAI also described GPT-5.5 as more efficient than GPT-5.4. Efficiency in this setting usually means the model can produce useful output with less compute, lower latency, or both, which matters for developers running repeated calls through an API. OpenAI’s public release did not present GPT-5.5 as a small patch; it was introduced as the company’s most capable model to date in this line.

The company’s focus on agentic coding reflects a broader shift in how large language models are being used. Instead of only generating snippets of code on request, agentic systems are expected to break down tasks, call tools, inspect results, and continue working with limited supervision. That makes reliability, persistence, and long-context handling more important than raw chat fluency.

OpenAI’s announcement also put a large number on ChatGPT’s scale. The company said ChatGPT now has 900 million weekly active users and 50 million subscribers. Those figures show how much of OpenAI’s business now depends not just on frontier models, but on the consumer and enterprise products built around them.

The API pricing is also part of the story. OpenAI said GPT-5.5 is available at $5 per 1 million input tokens and $30 per 1 million output tokens. For developers, token pricing is the basic unit that determines cost when sending prompts into a model and receiving generated text back out.

For teams building software agents, these details matter because model choice affects cost, speed, and how much work can be done in a single session. A model with a larger context window can keep more source code, logs, or instructions in memory, while a more efficient model can reduce the cost of running repeated tasks at scale.

OpenAI’s release also underscores how quickly the top end of the model market is moving. The company is no longer just competing on benchmark scores or demo quality, but on whether a model can stay useful inside real workflows, especially ones that involve coding, tools, and long-running automation jobs. GPT-5.5 enters that contest with a larger context window, a lower-friction API price point, and OpenAI’s biggest user base behind it.