OpenAI has moved its GPT-5.6 model family to general availability inside the ChatGPT consumer app, so the model you pick when you start a chat is changing again. The three-model line-up — Sol, Terra and Luna — began rolling out globally on July 9, 2026 and reached full availability within about a day. It replaces GPT-5.5 as the default choice for most people, though GPT-5.5 stays available for now. The older GPT-5.4 will be retired on July 23.
This follows last week’s GPT-Live voice mode rollout, another sign of how quickly OpenAI is shipping consumer-facing updates.
What the three GPT-5.6 models are
OpenAI named the tiers after the sun, earth and moon, and the logic maps neatly onto how capable each one is:
- Sol — the flagship. OpenAI says it “sets a new standard for both intelligence and efficiency” across coding, knowledge work, cybersecurity and science. It also has an “ultra” mode that coordinates several agents in parallel for complex jobs (Pro and Enterprise only).
- Terra — a balanced model for everyday work. OpenAI positions it as competitive with GPT-5.5 but roughly half the cost to run.
- Luna — the fastest and lowest-cost tier in the family.
Who gets GPT-5.6, and which model
Here is the part that matters when you open the app. In a normal ChatGPT conversation, ChatGPT Plus, Pro, Business and Enterprise users can select GPT-5.6 Sol from the model picker’s medium and higher reasoning-effort settings. Pro and Enterprise users also get a “Sol Pro” option for the toughest tasks.
Free and Go users get GPT-5.6 Terra instead of Sol. Terra and Luna are not directly selectable in a standard chat — they appear specifically inside ChatGPT Work and Codex.
For Indian readers: ChatGPT Go, OpenAI’s budget tier at roughly ₹399/month, gets Terra rather than the flagship Sol. That means Go subscribers get the “balanced” model, while Plus and above unlock Sol. None of the consumer subscription prices have changed with this release. The only pricing OpenAI published is its API/developer token pricing (Sol at $5 input/$30 output per million tokens, Terra at $2.50/$15, Luna at $1/$6) — that is for developers, not a ChatGPT subscription change.
What OpenAI claims GPT-5.6 can do
These are OpenAI’s own figures, not independently verified benchmarks. The company says Sol posts new state-of-the-art results on BrowseComp (92.2%) and OSWorld 2.0 (62.6%), and that it surpasses Anthropic’s Claude Opus 4.8 on OSWorld while using about 85% fewer output tokens. It also reports a large jump on its internal cybersecurity exploit benchmark (73.5% versus GPT-5.5’s 47.9%), which it frames as both a capability and a safety concern.
OpenAI also highlights better “design judgment”: when asked to update a presentation, Sol is meant to follow the existing deck’s layout, typography and formatting rather than dropping elements, as the example slide above shows.
ChatGPT Work, launched alongside it
OpenAI also launched ChatGPT Work on July 9, a new agent tool built on GPT-5.6 that completes multi-step tasks across web, mobile and desktop using a built-in browser and your connected apps and files, and can be scheduled to run on its own. It is available immediately for Pro, Enterprise and Edu subscribers, with Plus and Business users getting access within the next few days. Inside Work and Codex, Free and Go users get Terra, while paid users can choose between Sol, Terra and Luna. OpenAI has also folded Codex into the ChatGPT desktop app on Mac and Windows.
How to check if you have GPT-5.6
No action is needed — the update rolls out automatically. To see it, open ChatGPT, tap or click the model picker at the top of a new chat, and look for “GPT-5.6.” Plus, Pro, Business and Enterprise users will find Sol under the higher reasoning-effort settings, while Free and Go users will see Terra.
