OpenAI announced GPT-5.2 on 11 December 2025, marking a significant model update [source]. The release introduces architectural changes and expanded training data, though the company has not disclosed full technical specifications or benchmark comparisons against GPT-5.1.

According to the changelog, GPT-5.2 includes "improved reasoning capabilities" and "enhanced context handling" across multiple domains. OpenAI states the model was trained on data through October 2025, extending the knowledge cutoff by several months from the previous version.

The update arrives without advance notice to enterprise customers or API users. Several developers reported on social media that production applications began routing to GPT-5.2 automatically, causing unexpected output variations in established workflows. OpenAI's API documentation confirms that requests to the "gpt-5" endpoint now default to the 5.2 variant unless version pinning is explicitly configured.

No deprecation timeline has been published for GPT-5.1. The company's model versioning policy permits automatic updates to non-pinned endpoints, a practice that has previously led to behavioural drift in deployed applications.

OpenAI's announcement includes sample outputs demonstrating performance on coding tasks and multi-step reasoning problems. Independent verification of these capabilities has not yet been conducted. The company has not released information about changes to safety filters, output formatting, or token pricing.

Users seeking to maintain consistency with GPT-5.1 behaviour must update API calls to specify "gpt-5.1" explicitly. OpenAI states that pinned versions remain available for a minimum of three months following a new release, though historical practice shows extended availability in most cases.