Key facts
- OpenAI launched its GPT-5.6 model family with variants Sol, Terra, and Luna.
- The company claims Sol is 54% more token efficient for AI coding tasks.
- GPT-5.6 is described as OpenAI's strongest cybersecurity model yet.
- A new tool, ChatGPT Work, was released for enterprise teams.
- OpenAI asserts its new models outperform Anthropic's offerings on the Artificial Analysis Coding Agent Index.
OpenAI has introduced its latest family of AI models, GPT-5.6, aiming to enhance its position in a competitive market. The new models come in three variants: Sol, the workhorse; Terra, an intermediate option; and Luna, a budget-friendly choice. CEO Sam Altman highlighted Sol's 54% improvement in token efficiency for coding tasks and emphasized GPT-5.6's advanced cybersecurity capabilities, noting that the Trump administration had previously expressed concerns about its potential misuse.
The company also launched ChatGPT Work, a desktop, web, and mobile application designed to assist enterprise teams with daily tasks such as drafting documents and presentations. OpenAI's announcement follows similar releases from competitors SpaceXAI and Meta.
OpenAI is positioning GPT-5.6 as a direct challenge to its primary rival, Anthropic. Citing the Artificial Analysis Coding Agent Index, OpenAI claims its Sol model surpasses Anthropic's Fable 5 by 2.8 points, while using less time and fewer tokens. The company also stated that Terra outperforms Fable 5, and Luna outperforms Opus 4.8.
The GPT-5.6 models are now available via ChatGPT, Codex, and the OpenAI API, with specific pricing per million tokens for each variant: Sol at $5 input/$30 output, Terra at $2.50 input/$15 output, and Luna at $1 input/$6 output.