OpenAI Launches GPT-5.6 Family with Sol, Terra, and Luna for Flexible AI Choices

Overview

OpenAI has publicized a limited preview of its GPT-5.6 model family, which contains three models (Sol, Terra, and Luna) that will allow users and developers to select based on intelligence, speed, and price.

The Three Models

GPT -5.6 Sol (the flagship model) is identified by OpenAI as the most powerful version released thus far, with its agentic gains in coding, biological, and cybersecurity activity.

GPT -5.6 Terra is a hybrid model that performs on par with GPT -5.5 daily at a 50% cost savings.

GPT -5.6 Luna is the fastest and least expensive of the three, designed for high-volume, lower-stakes activity.

Pricing, based on 1 million tokens: Sol is priced at $5 for input / $30 for output; Terra at $2.50 for input / $15 for output; and Luna at $1 for input / $6 for output.

Availability

OpenAI will begin with a limited preview of Codex and the API, restricted to selected trustworthy partners by request of the U.S. Government. In the next several weeks, more general access to the Codex, API, and ChatGPT will be provided. OpenAI has expressed that this form of government-controlled access should not become the long-term standard.

Safety and Features

These models will launch with what OpenAI describes as the most comprehensive safety stack ever assembled. OpenAI has increased safety protections against high-risk cyber activities and repeated misuse, and will release real-time classifiers that temporarily pause generation for review if they detect such immoral behavior.

According to current state-of-the-art benchmarks for command-line workflow performance, GPT-5.6 Sol sets a new standard on Terminal-Benchmark 2.1, achieving superior scores in GeneBench v1 for genomics & quantitative biology disciplines, as well as performing phenomenally on the ExploitGym cybersecurity benchmark. As reasoning effort increases in ExploitGym, all of the platforms (i.e., Sol, Terra & Luna) have made noticeable improvements in their abilities to complete tasks. Sol did not exceed OpenAI’s Preparedness Framework for cyber risk assessment (Critical threshold).

This story was also published on my newsletter called Spyrigend.

Tip Jar

Enjoy the article? Consider leaving a tip to support my writing efforts. All donations are greatly appreciated!

Bitcoin: 3BcFDV3DYHT8rZurPWUF9kZUkqSFQ9kfAS

Etherium: 0x8AD199f90bf118a8ed708C73AA199EF7EF5444ba

Solana: A8cxXWcs6ARYNzgjK2nN4jkgh8Xx7a5EJFt8JeTRZ2rv