On July 9, OpenAI officially launched the new GPT-5.6 model series, which includes the main version Sol, the intermediate version Terra, and the economical version Luna, aiming to regain industry leadership with higher efficiency and cost advantages in the increasingly fierce AI competition. The series is now available on ChatGPT, Codex, and the OpenAI API simultaneously.

As a core highlight of this release, GPT-5.6 is positioned as the most powerful cybersecurity model in OpenAI's history. In the context of the Trump administration's previous attempts to restrict its promotion, the model fully supports defensive security activities, covering key areas such as threat modeling, code review, vulnerability patching, and blue team exercises, achieving a balance between cutting-edge performance and compliance security.

OpenAI, ChatGPT, artificial intelligence, AI

In terms of the developer ecosystem and enterprise market, OpenAI has shown strong aggressiveness, directly competing with the recent strong competitor Anthropic in the enterprise segment. In the authoritative AI analysis coding agent index (ACI) benchmark test, Sol, hailed as "the best coding model so far," scored 80 points, setting a new industry record, surpassing Anthropic's recently released Fable5 by 2.8 points. More disruptive is that when producing the same results, Sol's output tokens and time are less than half of Fable's, with only two-thirds of the cost.

Additionally, the mid-range model Terra outperforms Fable5 slightly, while the economical model Luna also defeated Opus4.8. According to CEO Sam Altman's plan, the core advantage of the GPT-5.6 model series lies in high efficiency and excellent cost-effectiveness, with Sol's token efficiency in AI coding tasks significantly increasing by 54%.

With the launch of the models, OpenAI also introduced a dedicated office assistant tool for enterprise teams - ChatGPT Work. This tool runs on desktop, web, and mobile platforms, efficiently handling daily document work, including automatically drafting documents, generating spreadsheets, and presentations, further expanding its ecological footprint in the B-side collaborative office field.

Currently, the GPT-5.6 series has announced clear commercial pricing: the per million input/output token prices for Sol, Terra, and Luna are $5/$30, $2.50/$15, and $1/$6, respectively. With major players like SpaceXAI and Meta releasing similar products intensively this week, the introduction of GPT-5.6 undoubtedly marks that the global large model competition has fully shifted from mere "parameter scale competition" to a deep contest of "scenario effectiveness and commercial cost-effectiveness."