OpenAI introduced the GPT-5.6 model family - ForkLog

# OpenAI introduces the GPT-5.6 family of models

OpenAI has opened limited access to GPT-5.6 Sol, Terra, and Luna via API and Codex for a small group of trusted partners. The company launched the preview at the request of U.S. authorities.

Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.https://t.co/OoM83SyISN

— OpenAI (@OpenAI) June 26, 2026

OpenAI expects to make GPT-5.6 Sol, Terra, and Luna publicly available in the coming weeks. Before the launch, OpenAI presented the models’ plans and capabilities in advance to the U.S. government. At the request of U.S. authorities, the initial phase is being conducted in the form of a limited preview, and information about the partners was handed to the government. OpenAI added that it does not consider this arrangement a long-term norm.

In the company’s lineup, Sol is called the flagship model, Terra is the “workhorse” model for everyday tasks, and Luna is a fast and inexpensive option. According to OpenAI’s assessment, Terra delivers performance comparable to GPT-5.5 and costs half as much. The company calls Luna the most affordable model in the lineup.

OpenAI also said that GPT-5.6 Sol is the company’s strongest model at the moment. The series includes a max reasoning mode, which gives Sol more time to deeply work through tasks, and an ultra mode that uses sub-agents to speed up complex work.

According to OpenAI, Sol set a new record on Terminal-Bench 2.1 for command-line tasks; on GeneBench v1 it achieved a result higher than GPT-5.5 with fewer tokens; and on ExploitBench it was competitive with Mythos Preview while using about one-third of output tokens.

Source: OpenAI. In tests on ExploitGym, the company claims that all three models improved results in cyber tasks as reasoning depth increased.

OpenAI stated that GPT-5.6 Sol, Terra, and Luna received the strongest safety stack in the lineup. According to the company, GPT-5.6 Sol does not cross the Cyber Critical threshold under the Preparedness Framework. In tests with Chromium and Firefox, the model found bugs and exploit primitives, but it was unable to autonomously create a full working exploit under the tested conditions.

To verify protection, OpenAI used more than 700,000 GPU-hours in A100-equivalent compute for automated red teaming aimed at finding universal jailbreaks. During the preview stage, multi-layered measures are in place, including restrictions at the model level, real-time checks, signals at the account level, monitoring, and enforcement measures. The company warned that some requests may be blocked or processed for longer due to additional verification.

OpenAI also explained the new naming scheme: the number indicates the model generation, and Sol, Terra, and Luna represent capability tiers. The price per 1 million tokens is $5 for input and $30 for output for Sol, $2.50 and $15 for Terra, and $1 and $6 for Luna.

In July, OpenAI plans to launch GPT-5.6 Sol on Cerebras at speeds of up to 750 tokens per second. Access at this stage will also be limited to select clients while the company expands capacity.

Recall that in early June, OpenAI confidentially filed with the U.S. Securities and Exchange Commission an S-1 registration statement for a potential initial public offering.