OpenAI's GPT-5.6 Sol, Terra and Luna Launch Publicly This Thursday

OpenAI is bringing GPT-5.6 to the public this Thursday, but rather than releasing a single model, it's launching a family of three: Sol, Terra and Luna. Each targets a different point on the cost-versus-capability curve, giving developers and businesses more flexibility in how they deploy the new generation of models. Sol sits at the top as the flagship model, offering the most advanced capabilities in the GPT-5.6 line. It also comes with what OpenAI describes as its most robust safety stack to date, suggesting the company has invested heavily in guardrails and alignment work for the model expected to handle the widest range of high-stakes tasks. Terra is positioned as a cheaper mid-tier alternative, with OpenAI claiming it matches GPT-5.5 performance at half the cost. That makes it a direct pitch to developers who found GPT-5.5 sufficient for their needs but are looking to cut spending without a downgrade in output quality. Luna rounds out the lineup as the fast, low-cost option built for high-volume workloads, where speed and price matter more than peak capability. Preview access to the GPT-5.6 family is already rolling out globally, ahead of Thursday's full public launch. This staged approach gives OpenAI a chance to gather feedback and iron out issues before the models are available to everyone. The three-tier structure mirrors a broader industry trend of splitting model releases by cost and performance, letting OpenAI compete across the market rather than offering a single one-size-fits-all product.