GPT-5.6 (Sol, Terra & Luna) Reaches General Availability

CodexView original changelog

OpenAI moved GPT-5.6 out of partner preview and into general availability on July 9, 2026, rolling it out across ChatGPT, the new ChatGPT Work, Codex, and the OpenAI API. The family ships in three tiers, Sol, Terra, and Luna, with Sol claimed to beat Anthropic's Fable 5 on coding benchmarks while using roughly half the tokens and costing a third less. GitHub simultaneously added the new models to GitHub Copilot, giving developers a second path to GPT-5.6 alongside Codex.

Featured Video

A video we selected to help illustrate this changelog


GPT-5.6 (Sol, Terra & Luna) Reaches General Availability

On July 9, 2026, OpenAI moved its GPT-5.6 model family out of its narrow ~20-partner preview (announced June 26) and into general availability across ChatGPT, the newly launched ChatGPT Work, Codex, and the OpenAI API, with a global rollout completing within 24 hours.

Three Tiers, One Family

GPT-5.6 ships in three variants: Sol, the flagship model built for the heaviest reasoning and coding workloads; Terra, a balanced mid-tier option for everyday work; and Luna, the fastest and cheapest of the three. Pricing per million tokens is $5 input / $30 output for Sol, $2.50 / $15 for Terra, and $1 / $6 for Luna.

Coding and Security Performance

OpenAI positions Sol as its strongest coding and cybersecurity model to date. On the Artificial Analysis Coding Agent Index, Sol reportedly scores 80.0, ahead of Anthropic's competing Fable 5 model, while using less than half the output tokens, taking less than half the time, and costing roughly a third less. OpenAI also describes Sol as its strongest cybersecurity model yet, citing frontier performance on threat modeling, code review, patching, and blue-team simulation tasks using significantly fewer tokens than prior models.

Where It Shows Up

GPT-5.6 is now selectable across ChatGPT (including the new ChatGPT Work product for enterprise teams), inside Codex, and through the OpenAI API. GitHub also shipped support for the new model family in GitHub Copilot on the same day, giving developers another path to the models inside their existing editor workflow.

Context

The general availability launch follows Codex CLI 0.143.0's Amazon Bedrock support for GPT-5.6, shipped a day earlier on July 8 while the models were still in partner preview, meaning Codex had early access to the new family through AWS before the broader public rollout landed.