OpenAI officially unveiled GPT-5.5 yesterday, April 23, 2026. It is being positioned as a significant step forward in "agentic" intelligence—models capable of planning, using tools, and executing complex, multi-step tasks independently with minimal human guidance.

Here is a breakdown of the latest updates regarding the release:

Core Capabilities

Agentic Performance: The primary focus is on "doing" rather than just "thinking." It is specifically designed for agentic coding, computer use, and scientific research. It is capable of planning workflows, using tools, checking its own work, and iterating until a task is complete.

Reasoning & Efficiency: Despite being a more capable model, OpenAI reports that GPT-5.5 maintains the same per-token latency as GPT-5.4. It handles large-system reasoning and ambiguity more effectively.

"GPT-5.5 Thinking": This mode allows the model to utilize more internal compute time to verify its assumptions before responding, leading to smarter, more concise outputs for complex problems.

Availability

Tiers: The model is currently rolling out to ChatGPT Plus, Pro, Business, and Enterprise users.

Variants:

GPT-5.5 (Standard): Optimized for general complex tasks and research.

GPT-5.5 Pro: Designed for the most demanding workflows with higher-quality outputs and reduced latency.

API: OpenAI has stated that API access is coming "very soon," with specific safety and security safeguards currently being finalized for scale.

Safety & Research

Robust Safeguards: OpenAI emphasized that this is their "strongest set of safeguards to date." This includes targeted testing for advanced cybersecurity and biology capabilities.

Self-Improvement Focus: Company executives have highlighted that GPT-5.5 is particularly adept at research-oriented tasks, including helping to improve versions of itself, which they describe as a meaningful step

toward AGI.