OpenAI launches its new family of models with GPT-5.6

On Thursday, OpenAI introduced its latest suite of AI models, GPT-5.6, which aims to redefine performance in the competitive AI landscape. This new release includes three distinct variants: Sol, Terra, and Luna, each tailored for different user needs and applications. The Sol model stands out as the powerhouse of the trio, while Terra serves as a mid-range option, and Luna is designed to be more accessible for budget-conscious users. OpenAI promises that these models will significantly enhance capabilities in various domains, including enterprise applications, software development, and scientific exploration. CEO Sam Altman emphasized the efficiency of these models, stating that Sol is 54% more efficient in token usage for coding tasks compared to previous iterations. Notably, GPT-5.6 is touted as the company's most advanced cybersecurity model to date, demonstrating exceptional performance with reduced token requirements. This aspect has garnered attention, particularly in light of past concerns from the Trump administration regarding potential misuse of such powerful AI tools. The cybersecurity features of GPT-5.6 include support for defensive measures like threat modeling, code analysis, and simulated attacks, allowing organizations to identify vulnerabilities before malicious hackers can exploit them. Additionally, OpenAI launched ChatGPT Work, a versatile tool designed to assist enterprise teams with routine tasks such as document creation, spreadsheet management, and presentation preparation, accessible on desktop, web, and mobile platforms. This announcement follows similar releases from competitors like SpaceXAI and Meta, but OpenAI appears to be particularly focused on outperforming Anthropic, its major competitor in the AI sector. Anthropic has gained traction by appealing to enterprise clients, establishing itself as a friendly alternative in the AI race. OpenAI claims that its latest models surpass Anthropic’s offerings in key benchmarks, specifically citing the Artificial Analysis Coding Agent Index. According to OpenAI, Sol is their most proficient coding model yet, achieving a score of 80—2.8 points higher than Anthropic’s Fable 5—while consuming less than half the output tokens and requiring significantly less time. The performance advantage extends to the other models as well, with Terra slightly edging out Fable 5 and Luna surpassing Opus 4.8. The GPT-5.6 models are now available through ChatGPT, Codex, and the OpenAI API, with pricing structured according to input and output tokens: Sol is priced at $5 input and $30 output, Terra at $2.50 input and $15 output, and Luna at $1 input and $6 output.