GPT-5.5 Arrives in Codex: Faster Agentic Coding with Expanded Browser Use
CodexOpenAI launched GPT-5.5, its most capable frontier model to date, rolling it out across ChatGPT and Codex for Plus, Pro, Business, and Enterprise subscribers. The model delivers measurable gains in agentic coding, achieving 82.7% on Terminal-Bench 2.0 and resolving 58.6% of real-world GitHub issues on SWE-Bench Pro in a single pass. Codex now benefits from GPT-5.5's expanded browser capabilities — enabling agents to interact with web apps, click through pages, test flows, and capture screenshots — while matching GPT-5.4's per-token latency at higher accuracy. A GPT-5.5 Pro variant with higher accuracy is available to Pro, Business, and Enterprise tiers.
Sources & Mentions
5 external resources covering this update
OpenAI releases GPT-5.5, bringing company one step closer to an AI 'super app'
TechCrunch
OpenAI announces GPT-5.5, its latest artificial intelligence model
CNBC
OpenAI Releases GPT-5.5: Faster, Smarter—And Pricier
Decrypt
GPT-5.5
Hacker News
OpenAI upgrades ChatGPT and Codex with GPT-5.5: 'a new class of intelligence for real work'
9to5Mac
GPT-5.5 Launches as Codex's Most Capable Frontier Model
OpenAI introduced GPT-5.5 on April 23, 2026, rolling the model out to Plus, Pro, Business, and Enterprise subscribers in both ChatGPT and Codex. Described by OpenAI as "a new class of intelligence for real work," GPT-5.5 arrives seven weeks after GPT-5.4 and reflects the company's rapid development cadence throughout 2026.
Agentic Coding Performance
GPT-5.5's gains are most pronounced in complex, multi-step coding workflows. On Terminal-Bench 2.0 — a benchmark measuring sophisticated command-line reasoning — the model scores 82.7%, ahead of Claude Opus 4.7 (69.4%) and Gemini 3.1 Pro (68.5%). On SWE-Bench Pro, which measures real-world GitHub issue resolution end-to-end in a single pass, GPT-5.5 reaches 58.6%.
Developers who tested the model early reported that GPT-5.5 has a significantly better understanding of the "shape" of a software system — grasping why something is failing, where the fix is needed, and what other parts of a codebase would be affected. One early tester noted that the model resolved in three minutes a problem that had held them up for four hours.
Expanded Browser Use in Codex
With GPT-5.5 powering Codex, the platform gains enhanced browser capabilities. Codex agents can now interact with web applications — testing user flows, clicking through pages, and capturing screenshots — making it meaningfully faster to iterate on frontend designs, web apps, and game prototypes. These capabilities complement the computer use features introduced in the April 16 Codex update, creating a more complete agentic environment across browsers, files, documents, and desktop applications.
Token Efficiency and Speed
GPT-5.5 matches GPT-5.4's per-token latency in real-world serving while performing better across nearly every benchmark measured. Notably, the model completes identical Codex tasks with fewer tokens than its predecessor — which partially offsets the higher API pricing ($5/$30 per million input/output tokens, approximately 2x GPT-5.4's pricing) for users running heavy workloads.
GPT-5.5 Pro
A GPT-5.5 Pro variant offering higher accuracy is available to Pro, Business, and Enterprise users. The standard GPT-5.5 is available to Plus subscribers and above. OpenAI noted that API availability was imminent, with the initial rollout limited to ChatGPT and Codex surfaces while safeguards for API-level deployments are finalized.
Availability
GPT-5.5 rolled out gradually to maintain service stability, starting with Pro and Enterprise accounts before reaching Plus tier users. The model powers Codex directly, so all Codex users on eligible plans benefit from the upgrade without any configuration change.