OpenAI has introduced GPT-6 Astra, its flagship multi-modal model developed through advances in pre-training and reinforcement learning. In parallel, OpenAI expanded the model lineup with GPT-6 Sol and GPT-6 Luna, targeted at high-throughput agentic workflows, frontend development automation, and complex code generation.
Model Capabilities: Astra achieves benchmark gains in cybersecurity evaluations (reaching 42.4% on ExploitGym) while optimizing token consumption relative to earlier model iterations.
Enterprise Focus: Features integrated execution sandboxes for front-end interface creation and scientific dataset plotting.
Source: OpenAI Index & Announcements