OpenAI launches GPT-5.2 with improved coding, reasoning, image comprehension, and multi-step project handling. Phased rollout begins now.

New Delhi: OpenAI has formally announced the launch of its new GPT-5.2 model, merely days after CEO Sam Altman allegedly issued a “code red” warning to staff. The company promotes the new model as substantially enhanced, asserting that it is superior in creating spreadsheets, developing presentations, writing code, interpreting images, comprehending lengthy contexts, utilizing tools, and managing complex, multi-step projects. Furthermore, OpenAI affirms that GPT-5.2 “surpasses industry experts in well-defined knowledge work tasks” across 44 distinct professions.

OpenAI Benchmarking performance

In benchmarks conducted by OpenAI, the new GPT-5.2 model demonstrates improved performance over its antecedent, GPT-5.1, across multiple assessments:

Coding: In SWE-Bench Pro, a comprehensive assessment of practical software engineering skills, GPT-5.2 achieved a score of 55.6 percent, representing an increase of approximately five percentage points compared to GPT-5.1.

Also read : Google Launches Emergency Live Video on Android – Real-Time Help During Crises

Abstract Reasoning: In ARC-AGI-1, a criterion designed to evaluate an AI system’s capacity for abstract reasoning akin to human cognition, the new system exceeded GPT-5.1’s score by over 10 percentage points.

OpenAI specifically states that the “Thinking” variant of GPT-5.2 demonstrates improved empirical accuracy in responses, asserting that it makes 30 percent fewer errors compared to earlier versions.

Practical Expectations and Competitive Landscape

Although the company has articulated its capabilities, its actual effectiveness in practice remains to be demonstrated. OpenAI’s most anticipated release of 2025, GPT-5, fell short of user expectations, with numerous users expressing dissatisfaction over perceived “inadequate responses” from the chatbot. On the LMarena website, where human users evaluate and rank the optimal outputs generated by AI systems, GPT-5.1 is presently positioned in sixth place, while Gemini holds the leading position. Models developed by Anthropic and Elon Musk’s xAI occupy the intermediate rankings between Google and OpenAI.

Availability and Pricing

OpenAI has initiated the phased deployment of GPT-5.2 across its Instant, Thinking, and Pro versions, beginning with users enrolled in paid plans (Plus, Pro, Go, Business, and Enterprise). The company advises users that if they do not immediately observe the new model, they may verify again later, as it is being deployed gradually. The current version, GPT-5.1, will be accessible for an additional three months prior to its discontinuation.