OpenAI is releasing the GPT-5.6 family (Sol, Terra, and Luna) for general availability across ChatGPT, Codex, and the API after a limited preview. The product post claims large jumps on named cyber evals, including ExploitBench at 73.5% versus GPT-5.5 at 47.9%, ExploitGym nearly doubling peak pass rate under a two-hour cap, and SEC-Bench Pro at 71.2% versus 45.8%. It describes layered safeguards (model-trained protections, real-time checks, monitoring, account enforcement, and a reasoning monitor) and states the models stay below a Critical threshold for cyber and biology. Higher defensive cyber capability is gated through Daybreak Trusted Access for Cyber, with identity verification and Advanced Account Security required for individuals. Benchmark numbers, safety conclusions, and the ten-times cyber-block claim are OpenAI assertions in this packet, with footnotes noting cyber evals used reduced safeguards and some latency or cost figures are simulated.