OpenAI has announced GPT-5.5 and rolled out the model to ChatGPT and Codex users on select paid plans. The publication focuses on coding, computer use, knowledge work, and research tasks that require prolonged 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.
According to OpenAI, GPT-5.5 is designed to plan, use tools, verify its work, and continue multi-stage coding tasks. The model can handle technical work such as implementation, refactoring, debugging, testing and validation.
Coding benchmarks
According to OpenAI, GPT-5.5 improves GPT-5.4 in three coding scores while using fewer tokens. On Terminal Bench 2.0, a command-line workflow test, the model achieved 82.7%.
Terminal Bench 2.0 measures whether models can perform command-line tasks that require planning, iteration, and tool usage. SWE-Bench Pro tests whether models can solve real GitHub problems.
On SWE-Bench Pro, which evaluates resolution of GitHub issues, GPT-5.5 scored 58.6%. It also outperformed GPT-5.4 on Expert-SWE, OpenAI’s internal assessment for longer coding tasks, with an average estimated human completion time of 20 hours.
Several early testers described stronger coding behavior in Codex. Dan Shipper, founder and CEO of Every, said that GPT-5.5 reproduced the type of system rewrite that one of its engineers ultimately chose for a post-launch problem, while GPT-5.4 could not.
MagicPath CEO Pietro Schirano said GPT-5.5 merged a branch with hundreds of frontend and refactor changes into a master branch that also changed, getting the job done in about 20 minutes. Michael Truell, co-founder and CEO of Cursor, said GPT-5.5 stayed on task longer and showed more reliable tool usage than GPT-5.4.
Token usage and latency
According to OpenAI, GPT-5.5 uses fewer tokens to complete the same Codex tasks. Lower token usage can reduce the cost of Codex tasks, especially when models need to read repository context, review logs, run tests, and rework multiple files.
The model also corresponds to GPT 5.4 per token latency when deployed in the real world. This means that GPT-5.5 achieved higher benchmark values compared to GPT-5.4 without increasing per-token latency.
Citing Artificial Analysis’s Coding Index, OpenAI said GPT-5.5 offers higher coding performance at half the price of some competing Frontier coding models. OpenAI has not provided full price comparison details for these external models in the material provided.
Changes in infrastructure
According to OpenAI, GPT-5.5 was co-developed, trained and deployed on NVIDIA GB200 and GB300 NVL72 systems.
Codex helped OpenAI teams test infrastructure ideas and identify optimizations that would be worthy of deeper technical work. One example was load balancing and partitioning heuristics that divided requests across accelerator cores.
Before GPT-5.5, requests were divided into a fixed number of blocks. Codex analyzed weeks of production traffic patterns and generated custom heuristic algorithms to improve partitioning and workload balancing.
OpenAI said the change increased the speed of token generation by more than 20%. The company did not specify whether the improvement applies to all GPT 5.5 traffic or to a specific portion of its serving stack.
Protective measures
GPT-5.5 went through OpenAI’s security and governance process, including readiness assessments, domain-specific testing, and targeted assessments for advanced biology and cybersecurity capabilities.
The model’s biological, chemical and cybersecurity capabilities are rated high under OpenAI’s Preparedness Framework. The model did not meet the company’s critical cybersecurity level.
OpenAI also applies stricter controls on higher-risk cyber activities and repeated abuse. Verified cyber defenders can apply for Trusted Access for Cyber, starting with Codex, to leverage less restricted models for approved defense work.
Availability and prices
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 ChatGPT for Pro, Business and Enterprise users.
In Codex, GPT-5.5 is available for Plus, Pro, Business, Enterprise, Edu and Go users. In Fast Mode, tokens are generated 1.5x faster for 2.5x cost.
API access is not available at startup. OpenAI said GPT-5.5 and GPT-5.5 Pro will be coming to the API “very soon.”
GPT-5.5 will be available through the Responses and Chat Completions APIs for $5 per 1 million input tokens and $30 per 1 million output tokens. Batch and Flex pricing is offered at half the standard API rate, while priority processing costs 2.5 times the standard rate.
GPT-5.5 Pro will also come to the API, priced at $30 per 1 million input tokens and $180 per 1 million output tokens. OpenAI said GPT-5.5 is more expensive than GPT-5.4 but uses fewer tokens in the codex for most users.
(Photo by Emiliano Vittoriosi)
See also: OpenAI offers free AI coding tools for open source maintainers
Want to learn more about AI and big data from industry leaders? Checkout AI and big data trade fair takes place in Amsterdam, California and London. The comprehensive event is part of TechEx and takes place alongside other leading technology events, click here Here for more information.
AI News is powered by TechForge Media. Discover more upcoming enterprise technology events and webinars Here.
