OpenAI has launched GPT-5.5, a significant update to its large language model, along with a suite of new products and features including ChatGPT Images 2.0, Workspace Agents, and a dedicated model for healthcare.
The GPT-5.5 and GPT-5.5 Pro API now support a 1 million-token context window, allowing the model to process and reason over vast amounts of text in a single pass. The new model also achieved a 58.6% score on SWE-Bench Pro, a benchmark for software engineering tasks, marking a substantial leap in performance for coding and developer workflows.
Accompanying the model launch is ChatGPT Images 2.0, powered by a new model called gpt-image-2. This update introduces native reasoning capabilities for image generation, enabling the AI to think step-by-step about visual composition. It renders multilingual text with improved accuracy and supports up to 2K resolution output.
OpenAI also previewed Workspace Agents, a new feature for Enterprise and Education plans that leverages the Codex architecture to provide persistent, agent-like assistance. These agents can maintain state across sessions and execute complex multi-step tasks on behalf of users.
In a move targeting the healthcare sector, OpenAI released ChatGPT for Clinicians, a specialized version of the AI assistant tailored for medical professionals. Alongside this, the company introduced HealthBench, a professional benchmark designed to evaluate AI performance on clinical reasoning and medical knowledge tasks.
Finally, OpenAI open-sourced the Privacy Filter, a 1.5 billion parameter model that runs locally on user hardware to redact personally identifiable information (PII) before data is sent to cloud APIs. This lightweight model aims to help organizations comply with data privacy regulations while still leveraging powerful AI services.
These announcements were made during a major release window spanning April 19-25, 2026.