TheModelDesk

OpenAI announced GPT-5.6 on July 9, 2026, introducing three model variants designed to compete in an increasingly crowded AI marketplace. Sol serves as the flagship workhorse model, Terra offers intermediate capabilities, and Luna provides a budget-friendly option. According to CEO Sam Altman, Sol achieves 54% greater token efficiency on coding tasks compared to previous versions, directly addressing cost and performance concerns.

The company emphasizes GPT-5.6's cybersecurity defenses, calling it the "strongest cybersecurity model yet." Capabilities include threat modeling, code review and automated patching, and blue teaming—simulating attacks on systems to identify vulnerabilities before adversaries exploit them. The Trump administration had previously sought restrictions on the model's rollout due to misuse concerns, though OpenAI frames the release as defensive-focused.

OpenAI also launched ChatGPT Work, a workplace tool for enterprise teams accessible across desktop, web, and mobile platforms, designed to automate clerical tasks like document, spreadsheet, and presentation creation. The company benchmarks Sol against competitors using the Artificial Analysis Coding Agent Index, claiming superior performance to Anthropic's recently released Fable model—a direct competitive salvo in OpenAI's ongoing battle with the perceived enterprise underdog.