GPT 5.5

Supported Functionality

Description

GPT-5.5 is OpenAI's professional reasoning and coding model, released on April 24, 2026. At launch, it was positioned as a frontier model for the most complex professional work, with a focus on coding, tool-heavy agents, grounded assistants, and long-context workflows.

Compared with prior models, it can reach strong results with fewer reasoning tokens and improves outcome-oriented execution, constraint preservation, tool selection, and argument accuracy. It also enhances default image detail, long-running state compaction, and output-style control, while the Responses API provides prompt caching, tool search, computer use, Hosted Shell, Apply Patch, Skills, MCP, and other capabilities.

Key Capabilities

• Complex Reasoning: Five reasoning-effort levels—none, low, medium, high, and xhigh—let developers balance quality, latency, and cost by task difficulty.
• Software Engineering and Coding: Strong planning, repository navigation, tool use, modification, and verification support multi-step, testable engineering tasks.
• Long-Running Agents: More precise tool selection and arguments on large tool surfaces, combined with compaction and state management, support sustained execution.
• Long-Context Retrieval: A 1,050,000-token context can hold large codebases, document collections, and multi-turn state, with up to 128,000 output tokens.
• Vision and Computer Interaction: Image input preserves more visual detail by default under auto and can pair with Computer Use for interface understanding and operation.
• Instruction Following and Execution: The model can work from goals, success criteria, and constraints to choose an execution path and turn product intent into concrete plans and results.
• Structured Professional Output: Structured Outputs and configurable text.verbosity support clear, direct, JSON Schema-compliant responses.

Technical Strengths