OpenAI unexpectedly released GPT-5.2 this morning, directly targeting the professional productivity of real-world applications and clashing head-on with Google Gemini 3. (Background: ChatGPT will support direct PayPal payments in 2026, the final piece of the OpenAI e-commerce empire puzzle.) (Background Supplement: A comprehensive look at the three major features of OpenAI's native browser 'ChatGPT Atlas' – can AI agents shake Chrome's dominance?) OpenAI launched its flagship model GPT-5.2 today (12) without prior warning, and the market has positioned this update as a direct counterattack against Gemini 3. The new model, which focuses on the 'economic value' of real workflows, differs from past models that emphasized conversational experience; CEO Sam Altman positions it as a digital employee that can clock in at any time. The new model targets 'economic value' and, to quantify this value, OpenAI has abandoned academic benchmarks in favor of its self-developed GDPval index. According to RD World analysis, GDPval covers 44 types of knowledge work, and GPT-5.2 Thinking performed better than or equal to human experts in 70.9% of tasks during testing, while the previous generation only managed 38.8%. This means that when handling Fortune 500 financial statements or common LBO models on Wall Street, GPT-5.2 can deliver results at less than 1% of human labor costs and at nearly 11 times the speed. It can read and understand hundreds of related tables at once; moreover, the model can directly convert data into presentation charts and deliver them instantly through Microsoft 365 Copilot, providing a plug-and-play productivity solution for enterprises. According to official descriptions, GPT-5.2 introduces three variants: Instant, Thinking, and Pro, with a core focus on proactive working capabilities. The new Tool Calling mechanism allows the model to automatically connect with external software, from asking for requirements all the way to delivering results. In real software engineering task testing with SWE-bench Pro, the Thinking version scored 55.6%, and the Verified version reached 80%. Triple Whale's CEO commented: "Integrating fragile multi-agent systems into a single large agent... it's like magic." Meanwhile, ScreenSpot-Pro testing shows that the model can analyze scientific charts and UI interfaces, paving the way for automation in enterprises. The release rhythm has sounded a 'red alert.' In just four months, OpenAI has successively released GPT-5, 5.1, and 5.2. Altman publicly stated: "We are in a red alert phase, and this state will last at least until January 2026." In the SWE-Bench Pro testing simulating software engineering tasks, GPT-5.2 Thinking set a record with a score of 55.6%, surpassing Gemini 3 and becoming the 'best engineering assistant.' However, Gemini 3 still holds advantages in GPQA Diamond and ARC-AGI tests, indicating that both sides are entrenched in their respective fortresses. OpenAI has chosen to prioritize strengthening programming and business logic capabilities that can directly generate revenue, temporarily putting academic-oriented tasks on the back burner. Long texts and agents, a qualitative change in the development ecosystem. GPT-5.2 supports a 256k context window, allowing it to parse an entire code library at once. Early partners Databricks and Cognition reported a 38% reduction in error detection rates, marking the first time AI agents have shown stability suitable for production environments. Enterprises are no longer just asking how to write code; they are allowing models to directly execute refactoring and debugging. With a new government set to take office in 2026, the focus of AI competition is shifting from pure technical scoring to the reshaping of the global labor market. For Wall Street and business owners, the question is no longer which model is smarter, but who can most quickly convert AI into profits on the balance sheet. Computing cost: The API pricing for GPT-5.2 is $1.75 per million tokens for input and $14 for output. GPT-5.2-pro is priced at $21 per million tokens for input and $168 for output. Related reports: A comprehensive look at the three major features of OpenAI's native browser 'ChatGPT Atlas' – can AI agents shake Chrome's dominance? Sam Altman discusses with a16z founder: OpenAI will make radical bets on infrastructure, with Sora as an important strategic tool. Altman in a frame with Pokémon: I hope Nintendo doesn't sue us… OpenAI's new Sora model ignites copyright battles. "OpenAI launches GPT-5.2! Aiming to replace professionals, with lower hallucinations, API cost summary." This article was first published on BlockTempo (the most influential blockchain news media).

#AI