OpenAI introduces RobustMAD, a benchmark for evaluating multimodal small language models' real-world robustness in anomaly detection. The study reveals promising capabilities of compact models but also critical robustness gaps. RobustMAD provides actionable guidance for designing next-generation industrial inspection assistants.

OpenAI, the creator of ChatGPT, is reportedly seeking a staggering $1 trillion valuation for its upcoming initial public offering (IPO), signaling immense confidence in its generative AI leadership. This ambitious target follows SpaceX's record-breaking IPO and sets a new benchmark for the burgeoning AI industry, attracting significant attention from investors and competitors alike.

Deutsche Telekom is aggressively pursuing an "AI-native" strategy, leveraging OpenAI's ChatGPT and API to overhaul operations from customer service to network management. This ambitious transformation aims to fundamentally redefine telecommunications delivery and decision-making, setting a new benchmark for enterprise AI adoption and efficiency.

A new arXiv study shows OpenEvidence's specialized clinical tool beats top general‑purpose models (Claude Opus 4.8, Gemini 3.1 Pro, GPT‑5.5) on 620 real‑world point‑of‑care questions. Physicians across 30 specialties rated the specialized tool higher on accuracy, utility, source quality, verifiability and completeness.

OpenAI introduces IMCBench, a benchmark for multimodal large language models in image-grounded medical conversations, evaluating safety, accuracy, and uncertainty in diagnosis. The benchmark tests eight models, with Claude Opus 4.6 achieving the highest overall score. The results highlight the need for multi-dimensional evaluation frameworks in medical AI.