OpenAI Officially Releases the "GPT-5.6" Family, Rolled Out in Three Tiers
OpenAI has released the “GPT-5.6” model family for general availability. It’s structured in three tiers — the top-tier “Sol,” the balanced everyday model “Terra,” and the low-cost “Luna” — and is available across ChatGPT, Codex, and the API.
Details
- Performance: Sol posted state-of-the-art results in coding, knowledge work, cybersecurity, and science. It scored 80 on the “Artificial Analysis Coding Agent Index” coding benchmark, 2.8 points ahead of the previous top score, while using under half the output tokens and roughly a third of the cost
- Reasoning modes: Adds two new modes — “max,” which explores more deeply, and “ultra,” which runs up to four agents in parallel to finish tasks faster
- Knowledge work: Improved template adherence and visual quality when creating presentations, documents, and spreadsheets
- Security: Vulnerability detection capability nearly doubled compared to the previous model, also supporting defensive use cases such as secure code review and threat modeling
Plans and pricing (per million tokens)
| Model | Input | Output | Main availability |
|---|---|---|---|
| Sol | $5 | $30 | ChatGPT Plus/Pro/Business/Enterprise (medium effort and above) |
| Terra | $2.50 | $15 | Available by default for free users of ChatGPT Work and Codex |
| Luna | $1 | $6 | Paid users can choose between Sol/Terra/Luna |
Prompt caching costs 1.25x on writes and gets a 90% discount on reads.
How to try it
- ChatGPT: Sol is available on Plus and above. Terra is available even to free users on ChatGPT Work and Codex
- API: All three models are accessible, along with new access to Programmatic Tool Calling and multi-agent features
- Rollout began in stages immediately after the announcement, with full availability expected within 24 hours