OpenAI launches GPT-5.6 with Sol, Terra and Luna tiers

OpenAI used July 9 to move GPT-5.6 from preview into general availability. The headline is not only a bigger flagship model: OpenAI is turning the release into a three-tier family for different cost and latency profiles. Sol is the top model, Terra is the middle tier for everyday work, and Luna is the cheaper fast tier.

What changed #

The company says GPT-5.6 Sol improves coding, knowledge work, cybersecurity and scientific tasks while using fewer tokens in several benchmark settings. The product angle is just as important: GPT-5.6 is available in ChatGPT, Codex and the API, so the model launch lands directly inside the tools where people already use agents.

The most interesting new knob is ultra, a high-capability mode that coordinates several agents in parallel for complex work. That makes the launch less like another chat model upgrade and more like OpenAI pushing multi-agent execution into normal product packaging.

Why it matters #

For users, the immediate question is price-performance: if Terra and Luna can cover a large share of GPT-5.5-style work, the cost of useful agents drops. For competitors, the signal is harsher. The frontier is moving from single-answer intelligence to systems that can plan, run tools, inspect output and keep working across longer jobs.