Over 100 fake citations slip through peer review at top AI conference
An analysis of nearly 5,000 accepted research papers from the AI conference NeurIPS 2025 uncovered more than 100 fabricated citations – despite review by multiple experts.
Baidu's new AI model Ernie 5.0, which processes text, images, audio, and video in a unified architecture, is now officially available. According to the LMArena ranking from January 15, 2026, Ernie-5.0-0110 scored 1,460 points, placing 8th globally and 1st among all Chinese models. That puts it on par with OpenAI's slightly older GPT-5.1 (High) and ahead of both Google's Gemini 2.5 Pro and Anthropic's Claude Sonnet 4.5. The next best Chinese model is GLM-4.7 from Zhipu AI. In the math category, Ernie 5.0 came in second worldwide, trailing only GPT 5.2 (High).
The LMArena ranking is determined from numerous anonymous pair comparisons in which users choose the better model answer.
Under the hood, the model uses a mixture-of-experts architecture with around 2.4 trillion parameters - but less than 3 percent of those are active for any given query. For now, the model is only available at ernie.baidu.com. Unlike previous releases, Baidu hasn't published any weights yet, and there's no technical report or detailed documentation available. The company's most recent open release was Ernie-4.5-VL-28B-A3B-Thinking, a model that can manipulate images during its reasoning process - for example, zooming in on text to read it more clearly.
Cursor's agent swarm tackles one of software's hardest problems and delivers a working browser
Building a web browser from scratch is considered one of the most complex software projects imaginable. All the more remarkable: Cursor set hundreds of autonomously working AI agents to exactly this task and after nearly a week produced a working browser with its own rendering engine.
Meta Superintelligence Labs has completed its first internal AI models, Chief Technology Officer Andrew Bosworth revealed at the World Economic Forum in Davos. Speaking with Reuters, Bosworth said the models are "very good," but there's still "a tremendous amount of work to do post-training." He didn't share specifics about what the models can do.
At an Axios event, Bosworth shared his broader take on AI development. He noted that for everyday queries, the improvements between model generations—like GPT-4 to GPT-5—are getting smaller. Specialized applications like legal analysis, health diagnostics, and personalization, however, continue to see significant gains. That's why he believes the industry's massive AI investments will pay off eventually.
Ollama, the popular software for running AI models locally, now supports image generation on macOS. The feature is still experimental, with Windows and Linux support coming later. Two models are available at launch: Z-Image Turbo from Alibaba's Tongyi Lab is a 6-billion-parameter model that creates photorealistic images and can render text in both English and Chinese. The recently released Flux 2 Klein from Black Forest Labs is the German company's fastest image model yet, available in 4B and 9B variants.
Terminals such as Ghostty or iTerm2 display previews directly.
The 4B version of Flux 2 Klein runs on standard graphics cards with at least 13 GB VRAM, such as an Nvidia RTX 3090 or 4070. The smaller version is available for commercial use, while the larger version is restricted to non-commercial applications. Generated images save directly to the current directory, and users can tweak image size, step count, and seed values. Additional models and image editing features are planned.
Following Microsoft's lead, OpenAI is reaching out to communities affected by massive AI infrastructure expansion. Through its "Stargate Community" program, the company promises that its AI data centers won't increase electricity prices for local residents. Microsoft made a similar pledge recently.
To deliver on this promise, OpenAI plans to fund its own energy sources, battery storage, and grid expansion. Each location will get a tailored plan based on local needs. The company also says it will better protect water resources and local ecosystems.
One year after announcing the Stargate project in January 2025, OpenAI says it has more than half of its 10-gigawatt capacity target for 2029 in the planning stages. The first site in Abilene, Texas is already training AI systems. Additional locations are under development in Shackelford County (Texas), Milam County (Texas), Dona Ana County (New Mexico), Port Washington (Wisconsin), Saline Township (Michigan), and Mount Pleasant (Wisconsin).
OpenAI starts selling ChatGPT ads, charges by views instead of clicks
OpenAI has started offering ad placements in ChatGPT to dozens of advertisers. Unlike Google or Amazon, the company is initially charging based on views rather than clicks.
Interactive demo shows AI models have opinions - and Grok really likes Elon Musk
A new interactive demo from nonprofit CivAI reveals how differently AI models respond to ethical and political questions – and why Grok likes Elon Musk more than Mahatma Gandhi.
OpenAI is rolling out age prediction in ChatGPT to identify when an account likely belongs to someone under 18, so the system can apply the right experience and safeguards for teens. The model analyzes behavioral patterns like usage times, how long the account has been active, and the age users entered at signup. When someone gets flagged as a minor, ChatGPT automatically enables safety features that block, among other things, graphic violence, sexual roleplay, depictions of self-harm, and content about extreme beauty standards.
Adults who get incorrectly flagged as minors can verify their age by taking a selfie through the Persona service. Parents get additional controls, including rest periods and notifications when the system detects signs of acute distress. The feature launches in the EU in the coming weeks. More details on OpenAI's help page.
