OpenAI Unveils GPT-5.5

OpenAI has unveiled GPT-5.5, a new intelligence model optimized for real-world tasks. This model goes beyond simple responses, focusing on Agentic capabilities that autonomously perform planning, tool use, and result verification.

Key Features and Performance:

• Enhanced Agentic Capabilities: Excels at performing complex and ambiguous multi-step tasks such as coding, online research, data analysis, and software operation.
• Efficiency and Speed: While the model's intelligence has increased, it maintains the same per-token latency as GPT-5.4, and also reduces token usage during Codex tasks, improving efficiency.
• Benchmark Performance: The model recorded performance surpassing existing and competing models on key metrics including Terminal-Bench 2.0 (82.7%) and OSWorld-Verified (78.7%).

Release and How to Use:

• It is currently being rolled out sequentially to ChatGPT (Plus, Pro, Business, Enterprise) and Codex users.
• The GPT-5.5 Pro version is available to Pro, Business, and Enterprise users.
• API service will launch soon, following a review of security and safety requirements.