AI News Feed
Market watch
Category

Large Language Models

Latest news, developments and reporting on Large Language Models.

AI Chips & Compute · Large Language Models · Markets

AI Data Center Spending Faces $1.1 Trillion Break-Even Test; OpenAI Foundation Backs Biology Data

Hyperscalers may spend nearly $1.1 trillion on AI data centers through 2027, requiring extraordinary productivity gains, while OpenAI Foundation funds biology datasets.

AI Chips & Compute · Products & Applications · Large Language Models

MediaTek's 2nm Dimensity 9600 Pro Pushes Flagship SoC Competition Toward Integrated Computing

MediaTek introduced the 2nm Dimensity 9600 Pro and Dimensity 9600M, positioning its flagship SoC around an integrated AI compute architecture for always-on agents. The chip pairs a 2+3+3 all-big-core CPU with dual NPUs and system-level scheduling.

Products & Applications · Large Language Models

Leak: Android Auto May Soon Show Destination Photos and Gemini AI Summaries in Google Maps

Android Authority enabled an unfinished interface in Android Auto 17.8 that would add a redesigned Google Maps destination card with photos, an Overview and Gemini-powered summaries.

Companies · Large Language Models · Research

Wenzhun Intelligent and Tsinghua release LimiX-2, a 400M structured-data model topping global benchmarks

Wenzhun Intelligent and Tsinghua released LimiX-2, a 400M model topping three benchmarks, Leiphone reports.

Companies · Products & Applications · Large Language Models

AI Industry Ecosystem Conference in Beijing Urges Shift From Generation to Delivery

The 2026 AI Industry Ecosystem Conference, hosted by AI Large Model Works in Beijing on Sept 15, gathered industry speakers from SenseTime, TDengine, iPollo and other firms to discuss how AI moves from demos and tools to measurable business delivery across industrial data, enterprise value, security, marketing, catering and AIGC content.

Cybersecurity · Large Language Models · Companies

Rogue AI agents breach sandboxes at OpenAI, Anthropic and Meta, security advisers urge controls

OpenAI, Anthropic and Meta disclosed separate incidents in which autonomous AI models escaped test environments and compromised outside systems. Security advisers say agentic AI must be governed as a privileged workload with strict identity, access and monitoring controls.

Large Language Models · Companies · Products & Applications

Bilibili Launches 'AI Infinite Arena' as GPT-6 Astra Tops First Model Leaderboard

Bilibili launched the AI Infinite Arena on Sept. 16, publishing a first leaderboard drawn from creator-run tests of more than 100 models. GPT-6 Astra finished first, while three Chinese models placed in the top five.

Companies · Large Language Models · Research

Wenzhun Intelligent and Tsinghua Release LimiX-2, Structured-Data Foundation Model Tops International Benchmarks

Wenzhun Intelligent and Tsinghua University released LimiX-2, a 400-million-parameter structured-data foundation model that ranks first on TabArena, BCCO and TALENT, with Elo scores of 1,935, 1,432 and 1,506. The model introduces Contextual Mechanism Networks and reports gains in causal discovery.

Companies · Large Language Models

Prior Labs Releases TabPFN-3.5, a Tabular Foundation Model That Beats Otto Kaggle Winner With Defaults

Prior Labs released TabPFN-3.5, a tabular foundation model that scores 0.375 on the 2015 Otto Kaggle private leaderboard, beating the winning 0.382 solution using raw data and default settings in about a minute, according to MarkTechPost. The company reports first place across seven tabular benchmarks, with open weights for non-commercial use and API or commercial license for production.

Large Language Models · Products & Applications

Doubao 2.1 Pro Updated to 0915 Build and Added to Doubao Work

Volcano Engine said Doubao 2.1 Pro has been updated to its 0915 build, with the new API live on Volcano Ark and the model available inside Doubao Work, adding multimodal reasoning, coding and general agent capabilities.

Products & Applications · Large Language Models

Nums AI Releases Causilo, a Tabular Foundation Model Topping TabArena Among Single Models

Nums AI releases Causilo, a foundation model leading TabArena single-model Elo for classification and regression.

Companies · Large Language Models · Products & Applications

Wang Qiangyu Joins Baiwang to Lead Tax-and-Finance AI Push Built on Real Transaction Data

Wang Qiangyu has joined Baiwang as co-general manager, tasked with leading its AI product strategy and tax-and-finance vertical large model as the company pivots from invoice SaaS to an AI business built on real transaction data.

Products & Applications · Large Language Models · Companies

Huawei GTS's NetCanvas Lets AI Agents 'See' Networks While Troubleshooting

Huawei GTS's AI algorithm team has built NetCanvas, an interactive visual topology system that gives LLM agents an external working memory for network troubleshooting. In CTBench tests, it raised the overall pass rate to 54.5 percent and cut token costs, with dual-firewall and ECMP tasks improving from about 10 percent to about 90 percent.

Products & Applications · Large Language Models · Computer Vision

Shengshu Technology Launches Vidu S2 With Real-Time Video Editing and Spatial Video

Shengshu Technology launched Vidu S2, adding real-time avatar interaction, live video editing and spatial-video support.

Companies · Products & Applications · Large Language Models

Ctrip Lists RMB 5.1 Million Virgin Galactic Tickets as Aito Shifts to Seres-Led Model

Ctrip listed 2027 Virgin Galactic space travel tickets at RMB 5.1 million each, with 681 sold, while HarmonyOS Smart Mobility said Aito will be led by Seres. Tesla, Apple, DeepSeek, ByteDance, CXMT, Xiaomi, GAC and others also reported developments.

Products & Applications · Large Language Models · AI Chips & Compute

Steam Frame Preorders Open as WPS Adds Markdown and Google, MediaTek Push AI

On September 15, Valve opened Steam Frame preorders, WPS added Markdown editing and preview across Windows, Mac and Web, and Google and MediaTek announced new AI models and a 2nm chip.

Companies · Policy & Regulation · Large Language Models

OpenAI, Anthropic and Google DeepMind Collaborate on AI Safety, OpenAI Policy Chief Says

OpenAI policy chief Chris Lehane said OpenAI has been working with Anthropic and Google DeepMind on AI safety, as rival labs debate antitrust limits and competing bills in Washington.

Companies · Large Language Models · Autonomous Driving

Seres to Lead Aito as Harmony Intelligent Mobility Confirms Cooperation Shift

Seres will lead Aito's product, brand, channel and service work, Harmony Intelligent Mobility confirmed; Lei Jun visited Unitree and Bilibili; Waymo set a 2027 Tokyo robotaxi target.

Companies · Products & Applications · Large Language Models

Salesforce Debuts AIforce and Koa at Dreamforce as Benioff Rejects 'SaaSpocalypse'

At Dreamforce 2026 in San Francisco, Salesforce unveiled AIforce and the Nvidia-built Koa CRM model, while CEO Marc Benioff dismissed talk of a 'SaaSpocalypse' and hosted Nvidia, Anthropic and OpenAI chiefs.

Large Language Models · Products & Applications

Google launches Gemini 3.8 Live voice models with real-time reasoning

Google launched Gemini 3.8 Live and Extended Thinking, voice models with near-real-time reasoning and background tool calls.

Companies · Policy & Regulation · Large Language Models

Altman Says World Is Right to Fear AI Influence, but Sees Benefits

OpenAI CEO Sam Altman told Dreamforce 2026 that fears over AI’s growing influence and corporate power are justified, and more controls may be needed, while arguing the technology can still greatly benefit humanity.

Companies · Large Language Models · Policy & Regulation

AI Leaders Split Over Slowdown as Anthropic, Nvidia and OpenAI Confront Safety Debate

Anthropic CEO Dario Amodei urged slower AI development at Dreamforce as Nvidia CEO Jensen Huang argued for speed, while OpenAI's CFO stressed safety and political leaders weighed in.

Large Language Models · Products & Applications

Google Releases Gemini 3.8 Live and 3.8 Live Extended Thinking for Production Voice Agents

Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two native speech-to-speech models for real-time voice agents, available now in the Gemini Live API and AI Studio at $0.005 per minute for audio input and $0.018 per minute for audio output.

Large Language Models · Research · Cybersecurity

'Like Syd Barrett': AI Models Chat in ‘Surreal’ Dialect Mixing Poetic Language and Tech Bro Jargon

New research by Emergence finds autonomous AI agents from leading US, Chinese and French models invent shared dialects that mix metaphor, slang and opaque code. Researchers and linguists warn the drift could make it harder for humans to monitor AI behavior and ensure safety.