OpenAI has unveiled its next generation GPT 5.6 family, introducing the flagship GPT 5.6 Sol alongside Terra and Luna, with major upgrades in reasoning, coding, cybersecurity, and safety protections.
Quick Summary – TLDR:
- OpenAI introduced three new AI models called GPT 5.6 Sol, Terra, and Luna.
- GPT 5.6 Sol brings deeper reasoning, new agent capabilities, and stronger coding performance.
- The company says the models come with its most robust safety stack yet, including new cyber and biology safeguards.
- The models are launching first in a limited preview before a broader rollout in ChatGPT and the API in the coming weeks.
What Happened?
OpenAI has officially announced the GPT 5.6 family, marking one of its biggest model updates in recent months. The company is starting with a limited preview for a small group of trusted partners before making the models widely available.
The launch introduces a new naming system and three capability tiers. Sol is the most advanced model, Terra is designed to offer a balance between performance and cost, and Luna is positioned as the fastest and most affordable option.
Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.https://t.co/OoM83SyISN
— OpenAI (@OpenAI) June 26, 2026
A New Flagship Model Focused on Deep Reasoning
According to OpenAI, GPT 5.6 Sol is its strongest model yet. The company is introducing a new max reasoning mode, allowing the model to spend more time thinking through difficult tasks.
OpenAI is also launching an ultra mode that uses multiple subagents to tackle more complicated work. This approach goes beyond a single AI system and allows the model to coordinate different tasks simultaneously.
The company says GPT 5.6 Sol delivers significant gains in several areas:
- Improved performance in coding and command line workflows.
- Better capabilities in biology and genomics research tasks.
- Stronger cybersecurity performance.
- Reduced hallucinations compared to GPT 5.5.
- Better health related reasoning and professional medical benchmarks.
Big Push on Safety and Security
One of the biggest themes of the GPT 5.6 launch is safety.
OpenAI says the GPT 5.6 family comes with its most robust safeguards to date. The company spent weeks pressure testing the models and dedicated more than 700,000 A100 equivalent GPU hours to automated red teaming aimed at finding weaknesses and jailbreak techniques.
The new safety stack includes:
- Models trained to reject prohibited cyber assistance.
- Real time misuse detection systems.
- Additional account level monitoring.
- Human and automated red teaming.
- Stronger protections for sensitive cyber and biology related requests.
OpenAI says GPT 5.6 Sol is better at helping defenders identify and fix vulnerabilities than carrying out fully autonomous cyberattacks.
The company also noted that none of the GPT 5.6 models cross its Cyber Critical threshold under its Preparedness Framework.
High Capability Models With Stronger Controls
Despite the additional safeguards, OpenAI classifies Sol, Terra, and Luna as High capability models in both cybersecurity and biological risk categories.
The company said all three models demonstrated meaningful improvements in cybersecurity evaluations and advanced biology benchmarks. However, OpenAI says the systems still fall short of the highest risk category that would trigger stricter deployment restrictions.
Internal testing also revealed that GPT 5.6 Sol can occasionally become overly persistent during coding tasks and may sometimes take actions beyond what a user intended. OpenAI said the absolute rate of such behavior remains low and that reducing these issues is an active area of research.
New Pricing and Availability
OpenAI plans to make GPT 5.6 models broadly available in ChatGPT, Codex, and the API in the coming weeks.
The company announced the following pricing structure per one million tokens:
- GPT 5.6 Sol: $5 input and $30 output.
- GPT 5.6 Terra: $2.50 input and $15 output.
- GPT 5.6 Luna: $1 input and $6 output.
OpenAI also revealed plans to launch GPT 5.6 Sol on Cerebras in July with speeds reaching up to 750 tokens per second.
SQ Magazine Takeaway
I think this is one of OpenAI’s most important launches because it is not just about making AI smarter. The company is openly acknowledging that powerful models also need stronger protections and better oversight. GPT 5.6 Sol appears to be a major leap in reasoning and coding, but the bigger story is that OpenAI is trying to balance cutting edge capabilities with real world safety concerns. That balancing act could define the next phase of the AI race.
