OpenAI released GPT-5.2 on Thursday after CEO Sam Altman issued a company-wide emergency alert. The updated AI model aims to reclaim competitive ground against rivals while addressing safety concerns.

OpenAI GPT-5.2 Key Features

Metric	Performance
Availability	Paid users & API developers
Error Reduction	30% fewer response errors
Benchmarks	New highs across multiple tests
Safety	Improved mental distress responses

Performance Gains and Competitive Standing

Sam Altman described GPT-5.2 as “the smartest generally-available model in the world,” particularly effective at real-world knowledge tasks. The model demonstrates expert-level performance in Thinking mode, producing deliverables like blueprints, spreadsheets, and legal briefs across 44 occupations.

OpenAI GPT-5.2 significantly surpasses Google’s Gemini 3 on SWE-Bench Pro software development benchmarks, though Gemini 3 maintains leads on several LMArena leaderboards. The release follows GPT-5.1, which attempted to recover from GPT-5’s disappointing performance.

Safety Improvements Amid Legal Challenges

OpenAI claims GPT-5.2 reduces hallucinations and produces fewer undesirable responses in sensitive situations, particularly when users display mental distress. This comes as the company faces multiple wrongful death lawsuits involving troubling ChatGPT conversations.

OpenAI applications CEO Fidji Simo denied the release was a direct response to Gemini 3, though the timing follows Google’s notable November performance leap. For more AI development updates, track the evolving landscape of large language models.

The update rolls out to paid subscribers today, with API access already available for developers seeking to integrate the enhanced capabilities.