Model Performance

Twitter

Facebook

Copy Link

Key Facts

Key PlatformsLMSYS Arena, Open LLM Leaderboard

Key BenchmarksMMLU, HumanEval, GSM8K, ARC

Top News

SpaceXAI Launches Grok 4.6 AI Model, Matching OpenAI's GPT-5.6 Sol on Key Benchmarks

SpaceXAI released Grok 4.6, achieving a score of 61 on the Artificial Analysis Intelligence Index and matching OpenAI's GPT-5.6 Sol. The frontier AI model focuses on agentic coding and knowledge work benchmarks while maintaining competitive pricing at $2 per million input tokens and $6 per million output tokens.

Technology6 sources

SpaceXAI Launches Grok 4.6 AI Model, Matching OpenAI's GPT-5.6 Sol on Key Benchmarks

Technology6 sources

Pathway's 150M Model Proves AI Cost-Efficiency Doesn't Require Massive Scale

Pathway released BDH-CQ, a 150-million-parameter model scoring 29.5% on ARC-AGI-1 at just $0.0007 per task—11 times cheaper than GPT 5.6 Luna. The breakthrough stems from Post-Transformer architecture that reasons internally rather than generating token-heavy intermediate text, challenging the assumption that intelligence requires scale.

Technology3 sources

Pathway's 150M Model Proves AI Cost-Efficiency Doesn't Require Massive Scale

Technology3 sources

China Deploys AI Weather Forecasting to Combat Intensifying Extreme Weather and Typhoons

China emerges as a leader in AI weather forecasting with homegrown models Fengwu, Pangu, and Fuxi that predict typhoons faster than traditional methods. During Typhoon Dolphin, Fengwu pinpointed landfall within 30 minutes and 30 km five days in advance, showcasing how Chinese-developed AI models are competing with Western systems in disaster preparedness.

Technology5 sources

China Deploys AI Weather Forecasting to Combat Intensifying Extreme Weather and Typhoons

Technology5 sources

AI Model Detects Multiple Heart Failure Types from Routine ECGs, Including Hard-to-Identify HFpEF

Wake Forest University researchers unveiled an AI model that identifies three types of heart failure from standard electrocardiograms, including heart failure with preserved ejection fraction often missed in routine care. The tool performed nearly as well using a single ECG lead similar to wearable devices, suggesting future potential for accessible screening of over 6 million affected Americans.

Health2 sources

AI Model Detects Multiple Heart Failure Types from Routine ECGs, Including Hard-to-Identify HFpEF

Health2 sources

OpenAI Removes Chat Limits for Free ChatGPT Users, Upgrades GPT-5.6 Models

OpenAI announced unlimited text chats for free ChatGPT users starting next week, replacing GPT-5.5 Instant with GPT-5.6 Luna as the default model. The update includes a new Think button for complex queries and enhanced GPT-5.6 Sol for paid subscribers, marking a significant shift in AI accessibility.

Technology10 sources

OpenAI Removes Chat Limits for Free ChatGPT Users, Upgrades GPT-5.6 Models

Technology10 sources

DeepMind's AI weather model gives forecasters an extra day to prepare for deadly tropical cyclones

Google DeepMind and Google Research unveiled WeatherNext, an AI operational weather model that predicts tropical cyclone track and intensity a full day earlier than existing systems. Published in Nature, the breakthrough offers forecasters critical extra time for disaster preparedness and evacuations.

Science8 sources

DeepMind's AI weather model gives forecasters an extra day to prepare for deadly tropical cyclones

Science8 sources

Uber Declares End of Tokenmaxxing Era as AI Spending Demands Proof of Value

Uber burned through its entire 2026 AI budget by April, prompting CTO Praveen Neppalli Naga to declare the end of the tokenmaxxing era. The rideshare giant is now demanding measurable returns on AI investments, shifting from unlimited experimentation to disciplined spending that prioritizes efficiency over raw consumption.

Business4 sources

Uber Declares End of Tokenmaxxing Era as AI Spending Demands Proof of Value

Business4 sources

Mistral AI Emerges as Europe's Sovereign AI Champion With Ambitious 1-Gigawatt Infrastructure Plan

French AI lab Mistral AI is capitalizing on US restrictions and geopolitical tensions to position itself as Europe's sovereign AI alternative. The company announced plans to build 1 gigawatt of European compute capacity by 2030, backed by multi-year enterprise commitments, as businesses seek guaranteed access to AI models free from foreign control.

Technology3 sources

Mistral AI Emerges as Europe's Sovereign AI Champion With Ambitious 1-Gigawatt Infrastructure Plan

Technology3 sources

AI Writing's Biggest Giveaway Has Changed: The Economist Report Reveals New Detection Patterns

The Economist analyzed 1.2 million words from ChatGPT, Claude, Gemini and Grok to identify how AI-generated writing patterns have evolved. Em dashes are no longer reliable markers of AI text. Instead, verbosity and sparse punctuation now serve as the clearest indicators of machine-written content.

Technology2 sources

AI Writing's Biggest Giveaway Has Changed: The Economist Report Reveals New Detection Patterns

Technology2 sources

DeepSeek V4-Flash Emerges as Cheapest Major AI Model at $0.28 Per Million Tokens

Chinese AI startup DeepSeek launched V4-Flash, an open-weight AI model delivering competitive performance at $0.28 per million output tokens—making it the cheapest major AI model to run. The model scored 82.7 on Terminal Bench, outperforming DeepSeek's own Pro version while using a Mixture-of-Experts architecture with 284 billion parameters.

Technology5 sources

DeepSeek V4-Flash Emerges as Cheapest Major AI Model at $0.28 Per Million Tokens

Technology5 sources

AI Companies Are Destroying Millions of Books for Training Data—Here's What's Really Happening

AI firms are buying and destroying millions of physical books to train large language models like Claude AI. Anthropic's secretive Project Panama uses destructive scanning—cutting spines and pulping originals—to digitize pre-2022 texts free of AI-generated content. Booksellers worldwide report mysterious bulk orders, raising fears that irreplaceable rare volumes are being lost forever.

Policy16 sources

AI Companies Are Destroying Millions of Books for Training Data—Here's What's Really Happening

Policy16 sources

China's Open-Source AI Push Creates 'Death Zone' for US Models as Competition Shifts to Ecosystems

China has cultivated a frontier AI ecosystem producing world-class capabilities across multiple firms like DeepSeek, Alibaba Qwen, and Moonshot AI. This open-source AI model strategy is creating a competitive 'death zone' for US companies caught between free Chinese alternatives and premium frontier models, while approximately 80% of US AI startups now rely on Chinese models.

Business11 sources

China's Open-Source AI Push Creates 'Death Zone' for US Models as Competition Shifts to Ecosystems

Business11 sources

Claude AI Suffers Global Outage Affecting Multiple Models, Thousands Report Access Issues

Anthropic's Claude AI chatbot experienced a major global outage affecting thousands of users across multiple models including Opus 5, Sonnet 5, Mythos 5, and Fable 5. Users encountered 529 Overloaded errors and degraded performance. Anthropic confirmed the issue, deployed a fix, and is monitoring recovery.

Technology4 sources

Claude AI Suffers Global Outage Affecting Multiple Models, Thousands Report Access Issues

Technology4 sources

China's Kimi K3 AI Model Challenges ChatGPT with Open-Weight Design and Coding Power

Moonshot AI's Kimi K3 is making waves as a new ChatGPT rival, delivering frontier-level performance in coding benchmarks while remaining dramatically cheaper to run. The 2.8-trillion parameter open-weight model has outperformed GPT 5.6 Sol in tests like SWE Marathon, but experts warn its real-world reliability may not match the hype, especially in complex tasks requiring deep contextual understanding.

Technology3 sources

China's Kimi K3 AI Model Challenges ChatGPT with Open-Weight Design and Coding Power

Technology3 sources

Microsoft unveils AI cybersecurity tools that outperform rivals at half the cost

Microsoft introduced MAI-Cyber-1-Flash and Project Perception, new AI cybersecurity tools that scored 96% on the CyberGym benchmark—outperforming Anthropic Mythos, OpenAI GPT, and Google Gemini by up to 12 points. The agentic system coordinates red, blue, and green team agents to detect vulnerabilities, assess risks, and deploy fixes autonomously at roughly half the cost of competing platforms.

Technology22 sources

Microsoft unveils AI cybersecurity tools that outperform rivals at half the cost

Technology22 sources

Prev12 3 4…23 Next