Model Performance

Twitter

Facebook

Copy Link

Key Facts

Key PlatformsLMSYS Arena, Open LLM Leaderboard

Key BenchmarksMMLU, HumanEval, GSM8K, ARC

Top News

Pope's AI Warnings Flagged as AI-Generated by Detection Tool With 99.98% Accuracy Rate

Pangram Labs' new Chrome extension flagged several posts from Pope Leo XIV's X account as likely AI-generated, including warnings about artificial intelligence itself. The AI detection startup claims a 99.98% accuracy rate, but the incident highlights ongoing debates about the reliability of such tools and the proliferation of AI-generated content across the web.

Technology2 sources

Pope's AI Warnings Flagged as AI-Generated by Detection Tool With 99.98% Accuracy Rate

Technology2 sources

NeoCognition raises $40M to build self-learning AI agents that specialize like human experts

NeoCognition, an AI research lab spun out of Ohio State University, has emerged from stealth with $40 million in seed funding to tackle a critical problem: current AI agents complete tasks successfully only 50% of the time. The startup is developing agents that learn like humans by building world models of their operating environments, enabling them to rapidly specialize in any domain rather than remaining unreliable generalists.

Startups3 sources

NeoCognition raises $40M to build self-learning AI agents that specialize like human experts

Startups3 sources

Chinese robots shift focus from stunts to tackling household chores in real homes

Chinese robotics firms are pivoting from showcase demonstrations to practical applications. X Square Robot deployed home-cleaning robots in Shenzhen households, charging $21.90 for 3-hour shifts. Despite being slow and clumsy, the company argues real-world deployment is essential for improvement, targeting a potential market worth 20% of GDP.

Technology2 sources

Chinese robots shift focus from stunts to tackling household chores in real homes

Technology2 sources

AI Chatbots Give Misleading Health Advice Nearly Half the Time, Major Study Reveals

A major audit of leading AI chatbots including ChatGPT, Gemini, and Grok found that nearly 50% of responses to health questions contain misleading or problematic information. With one in four Americans now using AI for health advice, the study published in BMJ Open highlights urgent public health risks and the need for stronger oversight of AI in healthcare.

Health5 sources

AI Chatbots Give Misleading Health Advice Nearly Half the Time, Major Study Reveals

Health5 sources

OpenAI launches ChatGPT Images 2.0 with text rendering that finally works

OpenAI released ChatGPT Images 2.0, a new AI image model that can accurately generate text in multiple languages and create complex visual content like infographics and marketing materials. The model includes thinking capabilities, web information access, and can generate up to eight images from a single prompt while maintaining visual consistency.

Technology28 sources

OpenAI launches ChatGPT Images 2.0 with text rendering that finally works

Technology28 sources

Multimodal AI model predicts PIK3CA mutations in breast cancer with cost-effective precision

Researchers from Hebei Medical University Fourth Hospital developed a multimodal AI framework that combines whole-slide pathology images with clinical data to predict PIK3CA mutations in breast cancer. The model achieved an AUC of 0.745 in internal testing and demonstrated strong generalizability across external cohorts, offering a practical alternative to expensive molecular testing for guiding PI3K-targeted therapies.

Science2 sources

Multimodal AI model predicts PIK3CA mutations in breast cancer with cost-effective precision

Science2 sources

AI systems judge people with systematic biases stronger than humans, study reveals

A Hebrew University study analyzing over 43,000 simulated decisions reveals AI systems exhibit systematic and predictable biases when evaluating people. While AI mimics human trust, it relies on rigid criteria rather than holistic impressions, leading to stronger biases in critical decision-making roles like job screening and creditworthiness assessment.

Science2 sources

AI systems judge people with systematic biases stronger than humans, study reveals

Science2 sources

OpenAI releases GPT-5.5, advancing toward AI superapp with enhanced coding and research tools

OpenAI unveiled GPT-5.5 on Thursday, calling it its smartest and most intuitive AI model yet. Released just seven weeks after GPT-5.4, the model brings significant improvements in agentic coding, knowledge work, and scientific research. President Greg Brockman says it's a step toward OpenAI's planned superapp that will unify ChatGPT, Codex, and AI browser capabilities.

Technology35 sources

OpenAI releases GPT-5.5, advancing toward AI superapp with enhanced coding and research tools

Technology35 sources

Tech Debt Holds Back AI Growth but Leaders Unlock 73% Digital Revenue, IDC Study Finds

An IDC study surveying 1,400 Asia Pacific IT leaders reveals a stark divide in AI readiness. While tech debt and legacy systems constrain most organizations, a cohort of leaders generates 73% digital revenue compared to just 20% for mainstream peers by investing in strategic modernization programs and modern data architecture.

Business2 sources

Tech Debt Holds Back AI Growth but Leaders Unlock 73% Digital Revenue, IDC Study Finds

Business2 sources

Pharos Network partners with Hong Kong university to explore AI in prediction markets

Pharos Network has launched a joint academic research project with the University of Hong Kong's FinTech Academy to study AI's role in prediction markets. Eight Master's students will conduct a three-month study using on-chain datasets, with promising projects fast-tracked into Pharos' $10M incubation program for real-world deployment.

Science3 sources

Pharos Network partners with Hong Kong university to explore AI in prediction markets

Science3 sources

Microsoft launches MAI-Image-2-Efficient, a 41% cheaper and 22% faster AI image model

Microsoft unveiled MAI-Image-2-Efficient, a cost-effective image generation model priced at $19.50 per million image output tokens—41% cheaper than its flagship. The faster AI image model delivers 4x high throughput efficiency per GPU and ships just weeks after the original, marking the company's rapid push to build an independent AI stack beyond OpenAI.

Technology3 sources

Microsoft launches MAI-Image-2-Efficient, a 41% cheaper and 22% faster AI image model

Technology3 sources

VueBuds bring visual AI to earbuds with tiny cameras, offering privacy-focused alternative to smart glasses

University of Washington researchers unveiled VueBuds, AI earbuds equipped with rice-grain-sized cameras that let users ask questions about their surroundings. The prototype achieves 83-84% accuracy in object recognition and translation while addressing privacy concerns through local processing and low-resolution imaging. Unlike smart glasses, VueBuds leverage a device people already wear daily.

Science4 sources

VueBuds bring visual AI to earbuds with tiny cameras, offering privacy-focused alternative to smart glasses

Science4 sources

AI Chatbots Give Misleading Medical Advice Half the Time, Multiple Studies Reveal

Multiple studies reveal AI chatbots deliver problematic health advice 50% of the time, with some platforms fabricating medical references at rates up to 34%. As one in three Americans now turn to AI for health information, researchers warn these tools lack clinical judgment and produce authoritative-sounding but potentially dangerous responses, especially when patient data is incomplete.

Health17 sources

AI Chatbots Give Misleading Medical Advice Half the Time, Multiple Studies Reveal

Health17 sources

China nearly erases US lead in AI as Stanford report reveals widening gap between experts and public

Stanford's 2026 AI Index Report shows China has nearly closed the performance gap with US AI models, narrowing the lead to just 2.7%. While the US invested $285.9 billion in AI in 2025, the flow of AI researchers moving to America has dropped 89% since 2017. Meanwhile, documented AI incidents rose to 362, and public anxiety about AI's impact on jobs and elections continues to grow.

Science8 sources

China nearly erases US lead in AI as Stanford report reveals widening gap between experts and public

Science8 sources

Anthropic overhauls Claude pricing as users report decline in model performance

Anthropic is transitioning Claude enterprise customers from flat-rate subscriptions to usage-based pricing, potentially tripling costs for some users. The shift comes as complaints about the AI model's performance have escalated sharply since February, with developers reporting quality degradation, quota exhaustion, and service outages that have turned a once-favored coding assistant into a source of frustration.

Business11 sources

Anthropic overhauls Claude pricing as users report decline in model performance

Business11 sources

Prev 1…456…19 Next