OpenAI on Thursday unveiled its latest family of models, bringing a novel set of advanced programs to an increasingly crowded field of AI offerings.
GPT-5.6 comes in three flavors: Sol (considered a workhorse), Terra (a more in-between option), and Luna (a budget-friendly option). These models expand users’ capabilities in a variety of fields – the company promises powerful capabilities for enterprise work, coding, and even scientific research.
CEO Sam Altman recently promised that his company’s latest models are orders of magnitude more effective and cheaper than previous versions says CNBC that Sol is a 54% more effective token when it comes to AI coding tasks.
Most importantly, the company calls version 5.6 its “strongest cybersecurity model yet, achieving pioneering performance with significantly fewer tokens.”
Indeed, there has been much buzz about the model’s cyber capabilities, as the Trump administration has previously sought to limit its deployment, ostensibly out of concern that the model could be abused. GPT-5.6 supports defensive activities including threat modeling, code reviewing and patching, and blue teaming (simulating an attack on your own systems to find vulnerabilities before real hackers do).
OpenAI has also released a novel tool called ChatGPT Workwhich appears to have been designed as a workplace companion for corporate teams, running across desktops, web and mobile devices, that can facilitate with everyday office tasks such as creating documents, spreadsheets and presentations.
The newly announced OpenAI family of models follows similar releases this week from competitors SpaceXAI and Meta.
However, GPT-5.6 and its accompanying marketing seem best designed to target OpenAI’s main adversary, Anthropic. Anthropic has managed to gain traction in an underdog position in the AI race by focusing on enterprise customers and, as a result, gaining an increasing share of support.
Not to be outdone, OpenAI quotes: Artificial analysis coding agent indexwhich is a noteworthy benchmark, and claims that its latest family of models outshines the Anthropic models in every respect.
OpenAI calls Sol its “best encoding model yet” and has pointedly compared it to Anthropic’s recently released (and very popular) Fable game. Using the Coding Agent Index, OpenAI claims that Sol “establishes a new state of the art of 80, 2.8 points above Fable 5, while using less than half the output tokens, taking less than half the time, and costing about a third less.”
He adds: “This advantage extends across the family: Terra performs slightly better against Fable 5, and Luna outperforms Opus 4.8.”
The company says that version 5.6 is now available on ChatGPT, Codex and OpenAI API. Availability per million tokens is priced as follows: Sol is $5 in/$30 out, Terra is $2.50 in/$15 out, and Luna is $1 in/$6 out.