OpenAI has released a new flagship AI model that strengthens coding, scientific research, knowledge work and long-horizon agent tasks, with performance on multiple benchmarks exceeding the previous generation, according to Jiemian News. The upgrade also cuts costs: cache read pricing is down 75%, total cost for typical workloads is about 25% lower than the prior model, and highly agentic tasks are up to 45% cheaper. The new model is available to all users, while a separate model with a higher safety level is being offered only to selected professional institutions.