OpenAI released GPT-5.2 Pro and Thinking models, boosting coding, enterprise tasks, long-context understanding amid competition with Google.
OpenAI has unveiled the GPT-5.2 family, including Pro and Thinking variants, designed for superior reasoning, coding, long-context processing, and productivity tools. This release follows an internal push to counter rivals like Google, addressing monetization while pushing performance frontiers. Benchmarks show strong results in web development and software engineering tasks, with GPT-5.2 ranking highly on arenas like LMArena WebDev[1][4].
The models target enterprise and professional use, with reported 80% on SWE-bench Verified, signaling a leap in practical coding reliability. This positions OpenAI strongly in the intensifying LLM race, where specialized capabilities are key differentiators[4].
Circuit-Sparsity, a companion technique, deactivates low-impact neural circuits, slashing FLOPs by 50-70% for efficient inference without major accuracy drops, aiding scalable deployment[1].
Source: ninjaai.com