Major Update to OpenAI's Flagship Model: GPT-5.2 Series Doubles Inference Speed, Price Remains Unchanged

OpenAI announced significant performance improvements to its latest flagship AI model, with GPT-5.2 and GPT-5.2-Codex now offering a 40% increase in reasoning speed for all API customers.

This update stems from optimizations to the company's reasoning stack, which handles model queries. Importantly, these improvements maintain the same model architecture and weights, meaning the underlying capabilities and training data remain unchanged—only the latency has been reduced to speed up response times.

"Same model. Same weight. The company emphasized in the announcement that the latency is lower, and highlighted that the upgrade applies uniformly across all API usage, without requiring any adjustments from developers.

This improvement follows the initial release of GPT-5.2 at the end of 2025, which was leading in areas such as professional work, long-term agent tasks, and complex reasoning benchmarks. GPT-5.2-Codex is a specialized variant optimized for software engineering and agent coding—with advantages in code refactoring, migration, terminal use, and cybersecurity applications—also benefiting from the speed improvements.

Developers and users have long considered response time a key factor in improving real-world productivity, especially in iterative coding workflows and interactive applications. The 40% acceleration is expected to unlock new use cases previously limited by high latency, enabling more seamless integration into tools, applications, and enterprise systems.

This update has received positive feedback in the developer community, with reports of significantly faster interaction speeds in coding environments and API calls. Despite the performance improvements, the price remains unchanged, maintaining cost-effectiveness for a large user base.

OpenAI continues to rapidly iterate on its model family, and the reasoning improvements reflect the ongoing effort to balance cutting-edge intelligence with practicality. The announcement did not mention any changes to the availability or deprecation schedule for previous versions.