OpenAI Launches Its New Family of Models With GPT‑5.6

OpenAI officially unveiled its newest generation of artificial intelligence models on Thursday, introducing GPT‑5.6 as its latest flagship offering into a rapidly evolving and increasingly competitive global AI landscape. The update brings a more structured, tiered lineup designed to serve a wider range of users from individual developers and small businesses to large enterprises and research institutions while promising major gains in efficiency, speed, and specialized capabilities.

GPT‑5.6 is released in three distinct variants, each tailored to different performance and cost requirements:

• Sol: Positioned as the primary “workhorse” model, built for high‑complexity tasks, enterprise workloads, and advanced problem‑solving;

• Terra: A balanced intermediate option, offering strong performance at a more accessible price point;

• Luna: The budget‑friendly entry‑level model, optimized for everyday use and lighter workloads without sacrificing core functionality.

According to the company, this expanded lineup delivers meaningful improvements across key areas including software development, business operations, and scientific research.

Greater Efficiency and Specialized Capabilities

In an interview with CNBC, OpenAI CEO Sam Altman emphasized that the new models represent a leap forward in resource efficiency, describing them as “orders of magnitude more cost‑effective” than previous generations. He highlighted that Sol in particular achieves 54% higher token efficiency specifically for coding tasks, meaning it can produce the same or better results using far fewer computational resources than older models.

One of the most heavily promoted features of GPT‑5.6 is its performance in cybersecurity an area the company calls its “strongest yet.” OpenAI states that the model reaches frontier‑level security capabilities while using significantly fewer tokens than competing systems. This strength has drawn considerable attention, as the Trump administration had previously signaled potential restrictions on its release, citing concerns over possible misuse for malicious hacking or offensive operations.

In response, OpenAI has clarified that the model is primarily optimized for defensive cybersecurity work. It supports use cases such as threat modeling, automated code review and vulnerability patching, and blue‑teaming the practice of simulating attacks on an organization’s own systems to identify and fix weaknesses before real adversaries can exploit them.

New Tools for the Workplace

Alongside the model updates, OpenAI introduced ChatGPT Work, a dedicated application designed to function as a comprehensive workplace companion for enterprise teams. Available across desktop, web, and mobile platforms, the tool is built to integrate directly into daily workflows, assisting with routine and administrative tasks such as drafting documents, building and analyzing spreadsheets, creating presentations, and summarizing large volumes of internal data. The goal is to reduce time spent on repetitive work and allow employees to focus on higher‑value strategic activities.

Competing in a Crowded Market

The launch of GPT‑5.6 comes amid a wave of new releases across the industry. Just this week, competitors including SpaceXAI and Meta rolled out their own updated model families, signaling an acceleration in the pace of AI development. However, OpenAI’s latest offering appears to be specifically engineered to challenge its closest rival, Anthropic.

In recent years, Anthropic has built a strong reputation as the “responsible alternative” in the AI race, gaining significant traction among enterprise clients and steadily increasing its market share. To counter this, OpenAI has published benchmark results from the Artificial Analysis Coding Agent Index a widely recognized industry metric claiming that GPT‑5.6 outperforms Anthropic’s latest models across the board.

The company draws a direct comparison between its top model, Sol, and Anthropic’s recently released and highly anticipated Fable 5. According to OpenAI’s findings:

“Sol sets a new state of the art at 80 on the index 2.8 points above Fable 5 while using less than half the output tokens, completing tasks in under half the time, and costing roughly one‑third as much.”

The performance advantage is said to extend throughout the entire lineup:

• Terra scores slightly higher than Fable 5;

• Luna outperforms Anthropic’s Opus 4.8 model.

Availability and Pricing

GPT‑5.6 is now accessible across OpenAI’s main platforms: the consumer and business interface ChatGPT, the code‑focused Codex, and the OpenAI API for developers and third‑party integrations.

Pricing is structured per million tokens, with clear distinctions between input and output costs:

• Sol: $5.00 per million input tokens / $30.00 per million output tokens

• Terra: $2.50 per million input tokens / $15.00 per million output tokens

• Luna: $1.00 per million input tokens / $6.00 per million output tokens

With this tiered pricing model, OpenAI aims to make high‑performance AI more accessible while maintaining profitability, positioning GPT‑5.6 as a flexible solution capable of competing on both capability and cost in the next phase of the global AI race.