OpenAI has introduced a new suite of advanced AI models, enhancing its offerings in a steadily evolving landscape of artificial intelligence solutions.
The latest release, GPT-5.6, is available in three distinct versions: Sol (the primary model), Terra (an intermediate choice), and Luna (the cost-efficient variant). These new models broaden the scope of applications available to users, with promises of enhanced capabilities in enterprise tasks, coding, and scientific research.
According to CEO Sam Altman, these new models are exceptionally efficient and cost-effective compared to their predecessors. He noted that Sol boasts a 54% improvement in token efficiency for AI coding tasks, as discussed in a recent interview.
A significant highlight is the company’s assertion that version 5.6 represents its most robust cybersecurity model to date, delivering superior performance while utilizing fewer tokens.
There has been considerable discussion surrounding the model’s cybersecurity features, especially since prior administrations showed concern regarding potential misuse. GPT-5.6 is geared toward defensive measures such as threat modeling, code review and patching, and blue teaming (a simulated attack on one’s systems to identify vulnerabilities before actual breaches occur).
OpenAI has also introduced ChatGPT Work, a dedicated tool intended to serve as a workplace assistant for teams. This application, accessible on desktop, web, and mobile platforms, aids in routine clerical tasks like creating documents, managing spreadsheets, and preparing presentations.
This announcement aligns with similar product launches from competitors SpaceXAI and Meta earlier this week.
However, it’s clear that GPT-5.6 is strategically positioned to compete primarily against OpenAI’s main rival, Anthropic. Anthropic has garnered attention as a friendly contender in the AI sector by concentrating on enterprise solutions and steadily gaining market traction.
In response, OpenAI references the Artificial Analysis Coding Agent Index—a recognized benchmarking tool—claiming that its latest models outperform those of Anthropic significantly.
OpenAI highlights Sol as its “best coding model yet,” contrasting it with Anthropic’s much-anticipated release, Fable. According to the Coding Agent Index, OpenAI asserts that Sol sets a new benchmark at 80, exceeding Fable 5 by 2.8 points, while using less than half the output tokens and taking significantly less time, costing approximately a third less in total.
Furthermore, the advantage is consistent across the entire lineup: Terra exceeds Fable 5, while Luna outperforms Opus 4.8.
The company notes that version 5.6 is now available across various platforms including ChatGPT, Codex, and the OpenAI API, with pricing for usage per million tokens as follows: Sol at $5 input and $30 output, Terra at $2.50 input and $15 output, and Luna at $1 input and $6 output.
