OpenAI releases new GPT-5.2 model to challenge Google, xAI, and Anthropic

New Delhi:

OpenAI has officially released its new GPT-5.2 model, just days after CEO Sam Altman reportedly issued a "code red" alert to employees. The company touts the new model as significantly improved, claiming it is "better at creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, using tools, and handling complex, multi-step projects". Furthermore, OpenAI asserts that GPT-5.2 "outperforms industry professionals at well-specified knowledge work tasks" across 44 different occupations.

Benchmark performance

In benchmarks highlighted by OpenAI, the new GPT-5.2 model shows better performance than its predecessor, GPT-5.1, in various evaluations:

• Coding: In SWE-Bench Pro, a rigorous evaluation of real-world software engineering, GPT-5.2 scored 55.6 percent, a gain of approximately five percentage points over GPT-5.1.
• Abstract Reasoning: In ARC-AGI-1, a benchmark testing an AI system's ability to reason abstractly like a human, the new system surpassed GPT-5.1’s score by more than 10 percentage points.

OpenAI specifically notes that the "Thinking" variant of GPT-5.2 is better at answering questions factually, claiming it produces errors 30 percent less frequently than previous versions.

Real-World expectations and competition

While the company has stated its capabilities, its real-world performance remains to be seen. OpenAI's biggest release of 2025, GPT-5, did not fully meet user expectations, with many complaining about "dumb answers" from the chatbot. On the LMarena website, where human users rate and vote on the best outputs from AI systems, GPT-5.1 currently sits in sixth position, while Gemini holds the top spot, with models from Anthropic and Elon Musk's xAI occupying the positions between Google and OpenAI.

Availability and pricing

OpenAI has begun the gradual rollout of GPT-5.2 (in its Instant, Thinking, and Pro variants), starting with users on paid plans (Plus, Pro, Go, Business, and Enterprise). The company informs users that if they don’t see the new model right away, they can check back later because it’s being rolled out gradually. The current version, GPT-5.1, will be available for three more months before it is retired.

ALSO READ: Time magazine selects 'Architects of AI' as its 2025 person of the year