OpenAI launches GPT 5.6 model family with Sol as its new flagship

OpenAI launched the GPT-5.6 model family, introducing three tiers—Sol, Terra, and Luna—across ChatGPT, Codex, and its API. Sol, the flagship model, is positioned for stronger agentic coding, knowledge work, and cybersecurity with improved cost efficiency. The rollout began globally and is expected to reach full availability within 24 hours.

OpenAI and ChatGPT logos are seen in this illustration taken, February 3, 2023. REUTERS/Dado Ruvic/Illustration OpenAI launches GPT 5.6 model family with Sol as its new flagship The new model lineup introduces Sol, Terra and Luna across ChatGPT Codex and the API, with OpenAI pitching stronger agentic coding, knowledge work, cyber defense, and lower cost per task. OpenAI is rolling out GPT-5.6 https://openai.com/index/gpt-5-6/ across ChatGPT, Codex, and its API, introducing a new three tier model family led by Sol, its latest flagship model. The lineup also includes Terra, a balanced model for everyday work, and Luna, the company’s lowest cost option in the GPT-5.6 family. OpenAI says the new naming system separates the model generation from durable capability tiers, giving users and developers clearer choices across intelligence, speed, and cost. The release follows a limited preview and expands access to GPT-5.6 across OpenAI’s consumer, developer, and enterprise products. The rollout is starting globally and is expected to continue toward full availability over 24 hours. OpenAI is positioning GPT-5.6 Sol as its strongest model for coding, knowledge work, cybersecurity, and science. The company says the model delivers better performance per dollar by completing more successful work with fewer tokens and lower estimated cost than previous frontier models. The company says Sol can coordinate tool use, inspect intermediate results, and refine outputs before returning finished work. In the API, Programmatic Tool Calling lets the model write and run lightweight programs in memory to filter intermediate data and decide next steps without sending every result back through the model. OpenAI is also introducing higher compute settings for more demanding tasks. Max gives GPT-5.6 more time to reason, check work, and revise outputs. Ultra goes further by coordinating multiple agents in parallel, with OpenAI describing the default setup as four agents working across separate workstreams before synthesizing the result. Coding is one of the main areas OpenAI is using to frame the launch. The company says GPT-5.6 Sol sets a new state of the art on the Artificial Analysis Coding Agent Index and also improves on Terminal Bench 2.1 and DeepSWE, benchmarks focused on command line tasks and long horizon engineering work in real codebases. OpenAI also says Terra performs above Claude Fable 5 on some coding agent measures, while Luna outperforms Claude Opus 4.8 at lower estimated cost. OpenAI is making a similar pitch for knowledge work. GPT-5.6 is designed to work across documents, spreadsheets, presentations, Slack, Notion, Microsoft 365, Google Drive, and other workplace tools. The company says Sol improves the quality of editable presentations, financial models, documents, and spreadsheets, especially when users provide reference files or templates. The launch also puts safety back at the center of OpenAI’s model strategy. GPT-5.6 is more capable in cybersecurity and biology than earlier systems, but OpenAI says the models do not cross its Critical threshold in either category. The company says its safeguards combine model training, real time checks, monitoring, account level enforcement, and access controls for higher risk capabilities. OpenAI’s system card says GPT-5.6 Sol is treated as High capability in cybersecurity, with Terra and Luna also reaching the High threshold, though with lower overall capability than Sol. The company says the goal is to preserve legitimate defensive work such as secure code review, patching, threat modeling, and vulnerability validation while applying tighter controls to serious misuse. Pricing for the API starts at $5 per 1 million input tokens and $30 per 1 million output tokens for Sol. Terra is priced at $2.50 input and $15 output, while Luna is priced at $1 input and $6 output. OpenAI is also adding more predictable prompt caching, including explicit cache breakpoints and a 30 minute minimum cache life. Disclosure: This article was edited by Estefano Gomez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .