OpenAI confirms GPT-5.6 rollout after delay: Sol, Terra and Luna arrive Thursday

OpenAI has confirmed that the new GPT-5.6 AI models will be released on Thursday, July 9. The company will introduce three different variants called Sol, Terra, and Luna. The announcement comes after OpenAI spent several weeks making the models available only to a limited group of trusted partners before planning a broader rollout.   Also Read: Want more reach on LinkedIn? Try these 8 ChatGPT prompts

The GPT-5.6 series is made for different types of users. Sol is the flagship model designed for the most demanding tasks, while Terra is built as a balanced option that offers strong performance at a lower price. Lastly, Luna is positioned as a faster and more affordable model for everyday work.  Also Read: Meta launches Pocket AI app that lets users create mini games with text prompts

OpenAI delayed the wider rollout, following discussions with the US government over the release of advanced AI models with stronger cybersecurity capabilities. OpenAI took to X to announce that GPT-5.6 Sol, Terra, and Luna will launch publicly on Thursday, while CEO Sam Altman also posted, “GPT-5.6 sol launches thursday! happy building.” 

Why the rollout happened in phases 

Before making the models widely available to everyone, OpenAI first opened limited access to a small group of trusted organisations and partners. 

As per the company, it shared details about the models and their capabilities with government officials before moving ahead with the broader rollout. The company also said it does not want this kind of government review to become the standard process for future launches. 

GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday.

We’re expanding preview access globally now. pic.twitter.com/Uk5HcfSc2e

— OpenAI (@OpenAI) July 8, 2026

What is new in GPT-5.6 

According to OpenAI, GPT-5.6 Sol is its most powerful model yet. One of the biggest additions is a new Max Reasoning mode, which allows the model to spend more time solving complex problems before generating a response. 

The AI model will also get an Ultra mode, which uses multiple sub-agents to work on complicated tasks instead of relying on a single AI model. OpenAI says this helps speed up more advanced workflows. 

The models are said to offer improvements in coding, biology, and cybersecurity tasks. As per the company, GPT-5.6 Sol achieved “state of the art” results on Terminal-Bench 2.1 for command-line coding workflows. On GeneBench v1, the model is said to have performed better than the GPT-5.5 while using fewer output tokens. 

According to OpenAI, Sol performs better on vulnerability research and defensive security tasks while remaining below the Cyber Critical threshold defined under its Preparedness Framework.  

Stronger safety measures  

OpenAI says GPT-5.6 launches with its strongest safety system so far. The company spent more than 700,000 A100-equivalent GPU hours on automated red-teaming to identify potential jailbreaks and weaknesses before release.  

The models also include multiple layers of protection, including built-in safeguards, real-time content checks, account-level monitoring, misuse detection and human expert testing.  

Availability and pricing  

GPT-5.6 Sol, Terra and Luna are scheduled to launch publicly on Thursday, July 9. During the initial phase, the models will be available through the API and Codex, with broader availability for ChatGPT and developers expected in the coming weeks.  

OpenAI has also announced pricing for the three models. GPT-5.6 Sol will cost $5 per one million input tokens and $30 per one million output tokens. Terra will be priced at $2.50 for input and $15 for output, while Luna will cost $1 for input and $6 for output.  

Add Techlusive as a Preferred Source

The company has also introduced improved prompt caching with GPT-5.6, including explicit cache breakpoints and a minimum cache life of 30 minutes. OpenAI has also confirmed that GPT-5.6 Sol will arrive on Cerebras in July, where it is is expected to deliver speeds of up to 750 tokens per second for select customers.