engblogs

GPT-5.2 delivers state-of-the-art long-context reasoning, robust agentic tool-calling, and professional-grade capabilities across coding, spreadsheets, and presentations, driving new benchmarks like GDPval and SWE-Bench to new heights.

A decade-long journey of pioneering AI research and deployment at OpenAI, highlighting breakthroughs, challenges, and the ongoing mission to develop beneficial AGI for all humanity.

Disney and OpenAI partner to enable fan-generated short AI videos featuring iconic characters on Sora, integrating advanced AI technologies while ensuring responsible usage and creative rights protection.

GPT-5.2 introduces major advances in AI capabilities for professional knowledge work, excelling in coding, long-context reasoning, vision, tool usage, and multi-step task execution with improved accuracy, speed, and reliability over previous versions.

GPT-5.2 Pro and GPT-5.2 Thinking bolster AI-assisted science and mathematics by stronger multi-step reasoning, consistency, and collaborative workflows, accelerating hypothesis testing, proofs, and discovery across disciplines, as shown in FrontierMath benchmarks and real-world case studies.

Podium leverages advanced AI agents, powered by GPT-5.1, to help SMBs increase revenue by automating lead capture, customer engagement, and service personalization, resulting in 300% AI-driven growth and improved operational efficiency.

An overview of GPT-5.2 updates highlighting continuity in safety measures and introducing GPT-5.2 Instant and GPT-5.2 Thinking variants.

GPT-5.2 enhances scientific and mathematical research through advanced reasoning, enabling faster exploration and more reliable insights while emphasizing human oversight.

The Agentic Commerce Suite streamlines selling via AI agents by simplifying product discovery, checkout, payments, and fraud protection through a single integration with Stripe.

NeurIPS 2025 highlighted the rise of environment-driven evaluation, data-centric AI, and principled rubrics as key trends shaping AI research and development in 2026.

MIT researchers developed a novel back-end transistor stacking technique using amorphous indium oxide and ferroelectric hafnium-zirconium-oxide to create ultra-compact, energy-efficient microelectronics with integrated memory, enabling faster computation and reduced power consumption for advanced AI and data-centric applications.

Litestream VFS enables instant, point-in-time SQLite queries directly from S3 object storage using a lightweight VFS plugin and LTX compaction for efficient, real-time backups and restores.