OpenAI introduces the GPT-5.6 model family

OpenAI for general use introduces the GPT-5.6 model family. The new generation consists of three variants: Sol as the primary and most powerful model, Terra as a balanced option for daily tasks, and Luna as the most cost-effective model.

According to the company, GPT-5.6 is designed to be more efficient than previous models in tasks related to coding, knowledge-based work, cybersecurity, and scientific research. OpenAI noted that the Sol version provides higher performance in complex tasks, consumes fewer tokens, and executes certain processes faster.

The new models also feature max and ultra modes. The max mode allows more time for deep analysis and result verification. The ultra mode uses multiple agents in parallel to help complete long-running or multi-step tasks faster.

GPT-5.6 is not limited to generating text or code. It also has the ability to verify final results, identify flaws in interfaces or documents, and fix them. OpenAI stated that the model can provide more accurate results when working with presentations, documents, and spreadsheets.

The new generation is being rolled out gradually via ChatGPT, Codex, and OpenAI API. In the ChatGPT Work service, the model can work with data from work environments like Slack, Notion, Microsoft 365, and Google Drive to create ready-made documents, presentations, or other materials.

Developers can access Sol, Terra, and Luna models via API. OpenAI has set pricing per 1 million tokens: $5 for input and $30 for output for Sol; $2.5 and $15 for Terra; $1 and $6 for Luna.

The company stated that the models underwent extensive safety testing before public release, including human red teaming, automated tests, and additional safeguards against malicious use.