OpenAI releases GPT-5.6 with three tiers, long-context API support, and ChatGPT Work

OpenAI announced GPT-5.6 on July 9, 2026, introducing a new model family across ChatGPT, Codex, and its API.

What OpenAI announced

OpenAI’s announcement describes GPT-5.6 as a three-tier release: Sol, Terra, and Luna. According to OpenAI, the models are available across ChatGPT, Codex, and the OpenAI API, with the company positioning the family around stronger reasoning, longer-context work, and more capable tool use.

OpenAI’s materials say GPT-5.6 includes new agentic and programmatic tool-calling capabilities. In practical terms, that means the models are designed to interact more directly with software tools and structured workflows, rather than only producing text responses. OpenAI also published benchmark results, long-context evaluations, and per-token pricing alongside the release.

API details and pricing

The OpenAI Developers API catalog lists GPT-5.6 Sol as the frontier model in the family. According to the API documentation, GPT-5.6 Sol supports a 1,050,000-token context window and up to 128,000 output tokens. The same documentation says the model supports reasoning tokens.

OpenAI Developers lists GPT-5.6 Sol pricing at $5 per 1 million input tokens and $30 per 1 million output tokens. Those published prices give developers a clearer basis for estimating costs when using the model for long documents, codebases, or multi-step tasks.

The large context window is one of the most concrete technical changes in the release. A million-token context window can allow a model to work across large collections of text or code in a single session, although actual performance will depend on task design, retrieval quality, and cost constraints.

ChatGPT Work and workplace documents

Axios reported that OpenAI broadly released GPT-5.6 with Sol, Terra, and Luna variants. Axios also said the company introduced ChatGPT Work, powered by GPT-5.6, for creating documents, spreadsheets, presentations, and other work products from connected context.

That framing suggests OpenAI is continuing to move ChatGPT beyond chat-style interactions and toward software that can assemble work artifacts. The Axios report indicates that ChatGPT Work can use connected context, though organizations evaluating it would still need to review privacy, data-access, and compliance controls before connecting workplace information.

Efficiency claims

CNBC’s published transcript of an interview with OpenAI CEO Sam Altman records Altman saying GPT-5.6 Sol is 54 percent more token efficient on agentic coding tasks. The same transcript records Altman saying GPT-5.6 Sol is “as good or better” than other top models.

Those are OpenAI’s claims through its chief executive, and they should be read in that context. Token efficiency can matter for developers because it may reduce cost or latency on complex coding tasks, but the practical impact will vary by workload, prompt structure, and integration.

Why it matters

The GPT-5.6 release is notable less for a single headline feature than for the combination of model tiers, API-scale context length, published pricing, and deeper tool-calling support.

For developers, the OpenAI Developers documentation provides concrete operating parameters, including context length, output limits, reasoning-token support, and per-token pricing. For business users, Axios’s reporting on ChatGPT Work points to OpenAI’s effort to package GPT-5.6 into document and productivity workflows.

OpenAI’s own announcement and documentation provide the most detailed technical claims, while Axios and CNBC offer outside corroboration of the launch and executive commentary. Together, the sources show a release aimed at both frontier model users and organizations looking to apply AI systems to longer, more structured work.