OpenAI has officially launched the GPT-5.6 family of large language models for general availability, introducing a tiered architecture designed to scale intelligence and efficiency across professional and developer workflows. The lineup features GPT-5.6 Sol as the flagship tier, GPT-5.6 Terra as a balanced option for everyday tasks, and GPT-5.6 Luna optimized for speed and cost. Early benchmarks indicate Sol establishes new state-of-the-art results in coding, cybersecurity, scientific reasoning, and long-horizon knowledge work. Crucially, the models achieve frontier performance while consuming significantly fewer tokens and operating at a fraction of the estimated cost of competing and previous-generation systems. Technologically, the update introduces the ultra inference setting, which coordinates up to four parallel AI agents by default to accelerate complex problem-solving without sacrificing accuracy. The release also features programmatic tool calling within the Responses API, enabling the model to autonomously write and execute lightweight scripts to manage intermediate data and tool interactions, thereby reducing latency and token overhead. Enhanced design judgment allows the system to generate polished, interactive interfaces and maintain strict adherence to corporate templates in presentations and documents. The models also demonstrate substantial gains in automating research and engineering tasks, with internal metrics showing a dramatic increase in research compute utilization during testing. OpenAI has deployed a multi-layered safety framework calibrated to each model capability tier. The system combines trained-in protections with real-time reasoning monitors, continuous monitoring, and account-level enforcement to mitigate misuse while preserving legitimate dual-use applications in cybersecurity and biological research. Independent testing suggests the models excel at defensive security tasks, such as vulnerability patching and secure code review, without reaching thresholds for autonomous offensive exploitation or biological synthesis. The safeguards have been stress-tested through extensive automated red teaming and expert collaboration, with the company noting a conservative initial rollout that blocks substantially more potentially harmful activity than previous versions while offering fallback routing for benign prompts. GPT-5.6 is rolling out globally across ChatGPT, Codex, and the OpenAI API over the next 24 hours. Pricing is structured at five dollars input and thirty dollars output per million tokens for Sol, two dollars and fifty cents and fifteen for Terra, and one and six for Luna. The update introduces deterministic prompt caching with explicit breakpoints and a thirty-minute minimum retention period. Developers can access multi-agent orchestration via a beta in the Responses API, while subscription tiers determine access to high-effort settings. The release marks a strategic shift toward maximizing utility-per-dollar in enterprise automation, positioning the family as a foundational infrastructure layer for accelerated engineering, research, and operational workflows.