OpenAI Launches GPT-5.6 Sol, Introducing New AI Model Tiers

OpenAI has publicly launched its new flagship model, GPT-5.6 Sol, along with two additional models named Terra and Luna. This release follows a two-week preview that involved around 20 trusted partners and was held back by the U.S. Department of Commerce. Notably, this release marks a new naming protocol for OpenAI's models, which now reflect different capability tiers instead of solely numerical designations.

GPT-5.6 Sol stands out in benchmarks, achieving a score of 91.9% on Terminal-Bench 2.1, a test that evaluates command-line workflows. In this ultra mode, Sol outperforms competitors such as Anthropic's Claude Mythos 5 and OpenAI's previous model, GPT-5.5. Priced at $5 and $30 per million input and output tokens, Sol lies between more expensive U.S. models and cheaper Chinese counterparts. Terra is described as matching GPT-5.5 at half the cost, while Luna provides an even lower-cost option.

Two new features accompany the launch: a max reasoning effort to extend Sol's cognitive capabilities and an ultra mode that utilizes subagents to optimize tasks. OpenAI's pricing strategy appears competitive, considering that Anthropic's Claude models are priced significantly higher.

According to OpenAI CEO Sam Altman, “With these new models, we aim to provide tailored AI solutions for different use cases, ensuring every user can find a model that fits their needs and budget.”

With this update, OpenAI continues to push the boundaries of AI performance while accommodating a broader range of users through tiered pricing. Future developments, including further enhancements and user feedback, will be important to monitor as the AI landscape evolves.