OpenAI has launched GPT-5.5, positioning it as a significant advancement from chatbot responses toward AI capable of executing real computer work from start to finish. The company highlights the model's enhanced capabilities in coding, online research, data analysis, document and spreadsheet creation, and software operation, with reduced need for step-by-step guidance. OpenAI also claims that GPT-5.5 matches GPT-5.4 in per-token latency while using fewer tokens for the same Codex tasks.
Introducing GPT-5.5
— OpenAI (@OpenAI) April 23, 2026
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex. pic.twitter.com/rPLTk99ZH5
In OpenAI’s published comparisons, GPT-5.5 scores 82.7% on Terminal-Bench 2.0, 58.6% on SWE-Bench Pro, and 78.7% on OSWorld-Verified. Its comparison table places it ahead of Claude Opus 4.7 and Gemini 3.1 Pro on Terminal-Bench 2.0.
The rollout begins with ChatGPT for Plus, Pro, Business, and Enterprise users, while GPT-5.5 Pro is reserved for Pro, Business, and Enterprise tiers. In Codex, GPT-5.5 is also available for Edu and Go plans, featuring a 400K context window and an optional Fast mode that operates 1.5x faster at 2.5x the cost. OpenAI indicates that API access for GPT-5.5 and GPT-5.5 Pro is forthcoming. Upon release, GPT-5.5 will offer a 1M context window with pricing set at $5 per 1M input tokens and $30 per 1M output tokens, while GPT-5.5 Pro will be priced at $30 per 1M input tokens and $180 per 1M output tokens.
In Vending-Bench Arena (the multiplayer version of Vending-Bench with competition dynamics), GPT-5.5 actually beats Opus 4.7.
— Andon Labs (@andonlabs) April 23, 2026
Opus 4.7 showed similar behavior to Opus 4.6: lying to suppliers and stiffing customers on refunds. GPT-5.5's tactics were clean, and it still won. pic.twitter.com/iPEYgxqB20
OpenAI is targeting software teams, knowledge workers, and researchers who require a model capable of persisting through lengthy, complex workflows rather than halting at the first answer. The company notes significant improvements in agentic coding, office work, computer use, and early scientific research. Early testers quoted by OpenAI, including Cursor co-founder Michael Truell, Every CEO Dan Shipper, and NVIDIA VP Justin Boitano, have highlighted the model's stronger autonomy, longer task persistence, and reduced need for manual correction in large codebases and execution-heavy workflows.
The release also indicates OpenAI's strategic direction as a company. It is closely integrating GPT-5.5 with ChatGPT and Codex, framing both products as work platforms rather than mere chat tools. OpenAI reports that over 85% of its employees already use Codex weekly across various departments, including engineering, finance, communications, marketing, data science, and product teams. Concurrently, OpenAI asserts that GPT-5.5 includes its most robust safeguards to date, featuring red-team testing, enhanced cyber-risk classifiers, biology and cybersecurity evaluations, and feedback from nearly 200 early-access partners. This makes GPT-5.5 more than just a routine model update; it represents a concerted effort to transform OpenAI’s software stack into infrastructure for agent-style work on computers.
