OpenAI has officially launched its new GPT-5.6 models for general availability, introducing a tiered family of artificial intelligence systems designed to balance performance and cost.

The new release features three distinct tiers tailored for different workloads. Sol serves as the flagship model, setting a new standard for intelligence and efficiency. Terra offers a balanced option for everyday tasks, while Luna provides a highly cost-efficient solution for high-volume applications.

Introducing Sol, Terra, and Luna

Developers can access these GPT-5.6 models through the API starting today. The flagship Sol model outperforms previous frontier systems in coding, knowledge work, and science. It achieves these results while using fewer tokens, which translates to a lower total cost for businesses. According to the OpenAI official announcement, the system delivers stronger performance per dollar, allowing teams to complete more work within the same budget.

“GPT-5.6 was the strongest model we evaluated on our agentic code-review tests. It beat GPT-5.5 while using roughly 3x fewer tokens.”

Itamar Friedman, Co-Founder & CEO at Qodo

Capabilities of the GPT-5.6 models

The new family excels in specialized technical domains. On the Agents Last Exam, which evaluates professional workflows across 55 fields, Sol achieved a score of 53.6, beating Claude Fable 5 by 13.1 points. In coding tasks, Sol set a new state-of-the-art score of 80 on the Artificial Analysis Coding Agent Index. It completed these tasks in less than half the time of competing models while reducing token usage significantly.

Advanced Multi-Agent Workflows

The safety safeguards built into the GPT-5.6 models are designed to prevent misuse. OpenAI implemented extensive testing, including 700,000 GPU hours of automated red teaming, to secure the models. For advanced tasks, a new “ultra” setting coordinates four agents in parallel to solve complex problems faster. This multi-agent capability is particularly useful for cybersecurity defense and complex software engineering.

Pricing and Global Availability

The models are priced per million tokens to accommodate different budgets. Sol costs $5 for input and $30 for output. Terra is priced at $2.50 for input and $15 for output, while Luna costs $1 for input and $6 for output. These options make advanced tools more accessible for developers building apps & software globally.