OpenAI’s new reasoning AI models, o3 and o4-mini, hallucinate more frequently than their predecessors, according to internal testing. Maxwell Zeff from TechCrunch reports that o3 hallucinated in 33% of questions on OpenAI’s PersonQA benchmark, approximately double the rate of previous models. The o4-mini performed even worse, with a 48% hallucination rate. OpenAI acknowledged in its …
The Washington Post has entered a strategic partnership with OpenAI to make its journalism accessible through ChatGPT. According to Todd Spangler’s article for Variety, the agreement will allow ChatGPT to display summaries, quotes, and links to Washington Post reporting in response to relevant search queries. The deal covers content across politics, global affairs, business, and …
A startup called Nari Labs has released Dia, a new open-source text-to-speech model designed to produce naturalistic dialogue. According to VentureBeat reporter Carl Franzen, the 1.6 billion parameter model rivals offerings from ElevenLabs, OpenAI, and Google’s NotebookLM. Co-creator Toby Kim developed Dia “with zero funding” and Google’s support through access to TPU chips. The model …
Google has released new versions of its Gemma 3 AI models that can run on consumer-grade graphics cards through a technique called Quantization-Aware Training (QAT). This development makes powerful AI models accessible to users without high-end hardware. The company announced that QAT dramatically reduces memory requirements while maintaining high quality performance. Gemma 3’s largest 27B …
Paid search engine Kagi has announced that its AI assistant feature is now available to all users across all subscription plans at no additional cost. According to Kagi’s announcement, the Assistant combines access to leading large language models (LLMs) with optional integration of Kagi Search results. The tool was previously exclusive to Ultimate subscribers but …
Google has released Gemini 2.5 Flash in preview, offering developers unprecedented control over the AI model’s reasoning capabilities. This new version allows users to toggle “thinking” on or off and set specific “thinking budgets” to balance quality, cost, and response time. The pricing structure reveals the cost impact of reasoning: input costs $0.15 per million …
Enterprise teams switching between large language models (LLMs) face numerous hidden challenges beyond simply changing API keys. According to an article by Lavanya Gupta, treating model migration as “plug-and-play” often leads to unexpected problems with output quality, costs, and performance. The report explores the complexities of moving between models like GPT-4o, Claude, and Gemini. Key …
In a wide-ranging interview at TED2025, OpenAI CEO Sam Altman addressed pressing questions about artificial intelligence safety, ethics, and the company’s future plans. The conversation, conducted by TED curator Chris Anderson, covered topics from OpenAI’s growth trajectory to the ethical implications of generative AI models. Altman revealed that ChatGPT has reached approximately 500 million weekly …
OpenAI has released a comprehensive prompting guide for its new GPT-4.1 family of models, highlighting significant improvements in coding capabilities, instruction following, and long context handling compared to GPT-4o. According to the guide published by OpenAI, developers may need to migrate their prompts because GPT-4.1 follows instructions more literally than previous versions, which tended to …
China’s ByteDance is transforming data from its popular apps TikTok, Douyin, and Toutiao into a competitive advantage in artificial intelligence, reports Meaghan Tobin of The New York Times. The company collects behavioral data from approximately 170 million U.S. TikTok users and around 1 billion users of its Chinese apps. This wealth of information has become …