OpenAI launches GPT-5.5 with better coding and lower token use

OpenAI has launched GPT-5.5, a new model it says is better at coding and more efficient to run than the GPT-5.4 release that arrived just last month. As first reported by The Verge, the rollout starts Thursday for Plus, Pro, Business, and Enterprise ChatGPT users, while GPT-5.5 Pro is limited to Pro, Business, and Enterprise accounts.

The release matters because OpenAI is clearly pushing beyond chatbot demos and deeper into day-to-day knowledge work. The company is positioning GPT-5.5 as a model that can take messy, multi-part assignments, plan the work, use tools, check results, and keep moving without constant hand-holding. That is a more ambitious pitch than simply answering questions faster.

According to OpenAI, GPT-5.5 is designed to handle writing and debugging code, web research, spreadsheets, and document creation more reliably. The company also says the model uses significantly fewer tokens in Codex, which matters for teams that are treating AI as part of a real software workflow rather than an occasional assistant.

The timing is important as well. OpenAI and Anthropic are now competing aggressively in AI coding, enterprise tooling, and cybersecurity-oriented models. A model that improves coding quality while lowering token consumption gives OpenAI a practical selling point for developers and larger organizations that watch both output quality and operating cost.

There is still a familiar caveat: model launch claims are not the same thing as independent benchmarking in production. But GPT-5.5 is still a meaningful release because it shows where the market is moving. The next phase of AI competition is less about flashy chat responses and more about whether these systems can complete useful work across tools with enough reliability to justify the bill.