OpenAI has announced GPT-5.5, with the model rolling out to ChatGPT and Codex users across selected paid plans. The release focuses on coding, computer use, knowledge work, and research tasks that require longer context and tool use.
In Codex, GPT-5.5 is available for Plus, Pro, Business, Enterprise, Edu, and Go users, with a 400K context window.
The larger context window is relevant for coding tasks that require access to more repository files, documentation, logs, or test output in one session.
GPT-5.5 is designed to plan, use tools, check its work, and continue through multi-step coding tasks, according to OpenAI. The model can handle engineering work such as implementation, refactoring, debugging, testing, and validation.
Coding benchmarks
GPT-5.5 improves on GPT-5.4 across three coding evaluations while using fewer tokens, according to OpenAI. On Terminal-Bench 2.0, a command-line workflow test, the model scored 82.7%.
Terminal-Bench 2.0 measures whether models can complete command-line tasks that require planning, iteration, and tool use. SWE-Bench Pro tests whether models can resolve real GitHub issues.
On SWE-Bench Pro, which evaluates GitHub issue resolution, GPT-5.5 scored 58.6%. It also outperformed GPT-5.4 on Expert-SWE, OpenAI’s internal evaluation for longer coding tasks with a median estimated human completion time of 20 hours.
Several early testers described stronger coding behaviour in Codex. Dan Shipper, founder and CEO of Every, said GPT-5.5 reproduced the type of system rewrite that one of his engineers had eventually chosen for a post-launch issue, while GPT-5.4 could not.
Pietro Schirano, CEO of MagicPath, said GPT-5.5 merged a branch with hundreds of frontend and refactor changes into a main branch that had also changed, resolving the work in about 20 minutes. Cursor co-founder and CEO Michael Truell said GPT-5.5 stayed on task for longer and showed more reliable tool use than GPT-5.4.
Token use and latency
GPT-5.5 uses fewer tokens to complete the same Codex tasks, according to OpenAI. Lower token use can reduce the cost of Codex tasks, especially when models need to read repository context, inspect logs, run tests, and revise multiple files.
The model also matches GPT-5.4 per-token latency in real-world serving. That means GPT-5.5 reached higher benchmark scores without increasing per-token latency compared with GPT-5.4.
OpenAI cited Artificial Analysis’s Coding Index, saying GPT-5.5 delivers higher coding performance at half the cost of some competing frontier coding models. OpenAI did not include full comparative pricing details for those external models in the material provided.
Infrastructure changes
GPT-5.5 was co-designed, trained, and served on NVIDIA GB200 and GB300 NVL72 systems, according to OpenAI.
Codex helped OpenAI teams test infrastructure ideas and identify optimisations worth deeper engineering work. One example involved load balancing and partitioning heuristics used to split requests across accelerator cores.
Before GPT-5.5, requests were split into a fixed number of chunks. Codex analysed weeks of production traffic patterns and generated custom heuristic algorithms to improve partitioning and workload balance.
OpenAI said that change increased token generation speeds by more than 20%. The company did not state whether the improvement applies across all GPT-5.5 traffic or to a specific part of its serving stack.
Safeguards
GPT-5.5 went through OpenAI’s safety and governance process, including preparedness evaluations, domain-specific testing, and targeted evaluations for advanced biology and cybersecurity capabilities.
The model’s biological, chemical, and cybersecurity capabilities are being treated as High under OpenAI’s Preparedness Framework. The model did not reach the company’s Critical cybersecurity capability level.
OpenAI is also applying tighter controls around higher-risk cyber activity and repeated misuse. Verified cyber defenders can apply for Trusted Access for Cyber, starting with Codex, to use less-restricted models for approved defensive work.
Availability and pricing
GPT-5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. GPT-5.5 Pro is rolling out to Pro, Business, and Enterprise users in ChatGPT.
In Codex, GPT-5.5 is available to Plus, Pro, Business, Enterprise, Edu, and Go users. Fast mode generates tokens 1.5 times faster for 2.5 times the cost.
API access is not available at launch. OpenAI said GPT-5.5 and GPT-5.5 Pro will come to the API “very soon.”
GPT-5.5 will be available through the Responses and Chat Completions APIs at US$5 per 1 million input tokens and US$30 per 1 million output tokens. Batch and Flex pricing will be offered at half the standard API rate, while Priority processing will cost 2.5 times the standard rate.
GPT-5.5 Pro will also come to the API, priced at US$30 per 1 million input tokens and US$180 per 1 million output tokens. OpenAI said GPT-5.5 is priced higher than GPT-5.4, but uses fewer tokens in Codex for most users.
(Photo by Emiliano Vittoriosi)
See also: OpenAI offers free AI coding tools to open-source maintainers
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.