智通财经网·2025-12-18 07:52Core Insights - The release of the GPT-5.2 series marks a significant transition from technical demonstrations to scalable economic production, showcasing AI's potential to create economic value in high-end professional fields [1] - GPT-5.2 has achieved a historic leap in core reasoning and professional task performance, reaching human expert levels in comprehensive assessments [1] Group 1: Model Performance - On December 12, OpenAI officially launched the GPT-5.2 series, which includes Instant, Thinking, and Pro versions tailored for varying task complexities [1] - In the ARC-AGI-2 test, known as the "Turing test for AI," GPT-5.2 scored 52.9%, nearly tripling the 17.6% score of GPT-5.1, and matched the abstract reasoning capabilities of the recently released Gemini 3 [1] - In the GDPval benchmark test covering 44 real-world job scenarios, GPT-5.2 Thinking outperformed or matched industry experts in 70.9% of tasks, while GPT-5.2 Pro achieved 74.1%, marking the first time an AI model reached top human levels in comprehensive knowledge work assessments [1] Group 2: Specialized Task Performance - In investment banking financial modeling tasks, GPT-5.2's average score improved from 59.1% to 68.4%, indicating deep penetration of AI into core productivity processes [1] - GPT-5.2 has shown significant advancements in code generation, long context handling, and visual understanding, providing reliable support for complex multimodal tasks [2] Group 3: Tool Reliability and Deployment - In the SWEBench Pro evaluation, GPT-5.2 Thinking achieved a 55.6% state-of-the-art score, demonstrating enhanced potential in front-end and 3D interface generation [2] - The long context processing capability has seen a qualitative leap, with nearly 100% accuracy in the "multi-needle retrieval" test at 256K token length, compared to only 30% for GPT-5.1, enabling deep analysis of lengthy documents and complex projects [2] - GPT-5.2's reliability in multi-step tool invocation tests (Tau2-bench) reached 98.7%, showcasing its strong end-to-end task execution capabilities [2] - OpenAI continues its iterative deployment strategy by offering the GPT-5.2 series to paid users in ChatGPT while retaining GPT-5.1 for three months to ensure a smooth transition [2]