GPT-5.5 (low) is a specific reasoning-effort configuration of OpenAI's flagship large language model series, officially released on April 23, 2026. It is designed to balance high-level cognitive performance with computational efficiency, serving as a more cost-effective variant of the standard GPT-5.5 model. This configuration is particularly optimized for agentic workflows, autonomous coding, and complex digital tasks that require multi-step reasoning without the high latency or token costs associated with "high" or "xhigh" effort tiers.
The model is built on an architecture that prioritizes "token efficiency," allowing it to complete complex tasks using significantly fewer tokens than the previous GPT-5.4 generation. In performance benchmarks, the "low" reasoning mode has demonstrated intelligence levels comparable to previous frontier models like Claude Opus 4.7 while operating at roughly half the cost. It excels in environments like Terminal-Bench 2.0 and Expert-SWE, where it handles long-horizon programming challenges by intuiting system structures and architectural dependencies with minimal human guidance.
Key technical features include a massive 1-million-token context window, enabling the model to ingest and reason over entire codebases or extensive technical documentation. GPT-5.5 (low) also integrates with OpenAI’s "Thinking" mode, which provides a brief overview of the model's planned reasoning approach before execution. This transparency allows users to interject or redirect the model's logic in real-time, making it more intuitive for professional research and development cycles.
As part of the broader GPT-5.5 ecosystem, this variant incorporates advanced cybersecurity safeguards under OpenAI's Preparedness Framework. While it remains a closed-weights model, it is a primary component of OpenAI's strategy for scalable AI agents, offering a versatile foundation for enterprises to deploy autonomous systems in software engineering, legal research, and data science.