GPT-5.2 is Here

Quick Overview

GPT-5.2 is a significant step forward, especially in instruction-following and code generation, but it is notably slower than GPT-5.1, with its Pro version being too slow for daily use and its Thinking mode being slow for most questions, although it excels in deep reasoning tasks and shows major improvements in visual and long-context understanding.

Key Points: GPT-5.2 Thinking is a meaningful step forward in instruction-following and complex task execution, particularly code generation, which is noted as significantly better than GPT-5.1. The Pro version of GPT-5.2 is considered incredibly impressive for deep reasoning but is too slow for daily use, often resulting in long thinking times that waste time. Performance benchmarks show GPT-5.2 Thinking beating or tying industry professionals on 70.9% of GDPVal knowledge work tasks, outperforming GPT-5.1 significantly in areas like ARC-AGI-2 (52.9% vs 17.8%). The model demonstrated a 390X efficiency improvement in one year on the ARC-AGI-1 task when comparing GPT-5.2 Pro SOTA score (90.5% at $11.64/task) to a previous version (88% at $4.5/task). Early feedback suggests GPT-5.2 excels in professional deliverables like spreadsheets and presentations, with the Pro version being more polished and deliberate than the chaotic freelancing style of GPT-5.1. The Disney/OpenAI partnership was announced, granting Sora access to Disney characters for user-prompted video generation, with Disney also investing $1 billion in OpenAI. A key downside noted is the speed; GPT-5.2 Thinking is slow for most questions, and Pro is often too slow for real-time use, suggesting it's best reserved for complex, deep reasoning tasks.

Context: The video aggregates reactions and initial reviews from various AI researchers and industry figures regarding the release of OpenAI's new model, GPT-5.2, which was announced alongside a major partnership with The Walt Disney Company for their Sora video generation model. The discussion centers on the improvements GPT-5.2 offers over its predecessor, GPT-5.1, particularly in professional tasks, reasoning, and coding, while also highlighting significant trade-offs, primarily in speed, especially with the more capable Pro version.