OpenAI Launches GPT-5.2 Amid Escalating AI Competition
OpenAI has launched GPT-5.2, its most advanced artificial intelligence model to date, marking a significant upgrade in capabilities for professional users and developers. The new model rolls out on December 12, 2025, across ChatGPT’s paid subscription tiers and through the company’s application programming interface.
GPT-5.2 introduces substantial improvements in workplace productivity tasks including spreadsheet generation, presentation building, code writing, and long-document analysis. The model is available in three variants: Instant for faster responses, Thinking for structured work requiring deeper reasoning, and Pro for maximum accuracy on complex queries.
Performance benchmarks demonstrate notable advances over the previous GPT-5.1 model. On GDPval, an evaluation measuring professional knowledge work across 44 occupations, GPT-5.2 Thinking achieves a 70.9 percent win or tie rate when compared against industry experts. The model completed these tasks at eleven times the speed and less than one percent of the cost of human professionals, though OpenAI emphasizes the importance of human oversight.
Software engineering capabilities show marked improvement with GPT-5.2 Thinking scoring 55.6 percent on SWE-Bench Pro and 80 percent on SWE-bench Verified. These benchmarks evaluate real-world coding tasks across multiple programming languages. Internal testing on investment banking analyst tasks showed a 9.3 percent improvement, with scores rising from 59.1 percent to 68.4 percent.
The model demonstrates reduced hallucination rates with errors decreasing by 30 percent relative to GPT-5.1 Thinking on ChatGPT queries. Long-context processing has improved substantially, with GPT-5.2 achieving near-perfect accuracy on tasks requiring information synthesis across hundreds of thousands of tokens.
Vision capabilities have advanced significantly, with error rates cut roughly in half for chart interpretation and software interface understanding. This enables more accurate analysis of dashboards, technical diagrams, and visual reports critical to business operations.
OpenAI has priced GPT-5.2 at 1.75 dollars per million input tokens and 14 dollars per million output tokens through its API, representing a 40 percent increase over GPT-5.1 pricing. The company maintains that improved token efficiency results in lower overall costs for achieving comparable quality levels.
The launch follows OpenAI’s declaration of a code red effort in response to competitive releases from Google and Anthropic. Google simultaneously released its Gemini 3 model and Deep Research agent, while Anthropic’s Opus 4.5 outperforms GPT-5.2 on certain coding benchmarks. OpenAI CEO Sam Altman indicated that competitive impact has been less severe than initially anticipated and expects to conclude the code red status by January 2025.
More than 800 million people use ChatGPT weekly, establishing it as one of the fastest-growing commercial platforms since its 2022 launch. The company operates with a 500 billion dollar valuation and has committed over 1.4 trillion dollars to planned infrastructure spending.
GPT-5.1 will remain available as a legacy option for three months before being phased out. The rollout to ChatGPT subscribers begins immediately for Plus, Pro, Team, and Enterprise users, with gradual deployment to maintain platform stability.
