Latest news, developments and reporting on Large Language Models.
Perplexity released pplx-embed-v2-late, ColBERT-style multimodal embedding models in 0.6B and 9B sizes, with MIT-licensed weights on Hugging Face. The 9B model scored 92.4% on MADQA, while the 0.6B model is designed for edge query encoding.
Huawei and Nvidia target KV cache storage; Ruqi builds real-world data infrastructure for physical AI, Leiphone reported.
CAIR released CARES 4.0, a multimodal clinical agent system, at the fourth CREATE Symposium in Hong Kong, where clinicians, hospital leaders, researchers and investors discussed medical AI's path into clinical workflows and commercialization.
US venture deal value reached a record $515.8 billion in the first nine months of 2026, lifted by OpenAI and Anthropic mega-rounds, while exits lagged far behind, PitchBook-NVCA data show.
Jieyue Terminal will launch STEPX Neo, its first large-model-native agent phone, in Shanghai on Oct 13.
Cisco used WebexOne to pitch a unified AI platform and argue context, token cost and trust will decide the agent era.
Anthropic launches Claude Haiku 5.5 as a low-cost small model; OpenAI rolls out GPT-6 to free ChatGPT users with new UI and a safety report.
Microsoft and Nvidia unveiled the RTX Spark-powered Surface Laptop Ultra, as Google, Anthropic and Reflection launched new AI models and tools.
CrowdStrike says a Chinese-speaking hacker used AI tools and LLMs to breach South Korean banks, Yonhap reports.
OpenAI says its GPT-6 models let ChatGPT decide when to replace text with graphics, tappable buttons, forms and charts. The Intelligent UI rolls out today to Plus, Pro, Business and Enterprise subscribers, and to free and Go users tomorrow.
OpenAI rolled out GPT-6 with an interactive UI; Apple, DeepSeek and Anthropic made major AI and device news.
Nvidia and Samsung joined a $90M Series B for Nous Research, maker of the Hermes AI agent, valuing it at $1.5B.
Google and Unity have announced Playground, a platform that lets users 18 and older generate and customize video games from natural-language prompts without coding. Unity Spark, a more advanced creation experience, is set to be integrated later in 2026.
Anthropic released Claude Haiku 5.5, a small model it says runs at roughly a quarter of Haiku 4.5's cost for most prompts, and halved cache-read prices on Claude Sonnet 5.5 as it competes with OpenAI's GPT-6 Luna.
Artcraft developer Brandon Thomas used Anthropic’s Claude Opus 5.5 to create seven open-source apps that mimic Adobe Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign and Acrobat Pro, Ars Technica reported. The software is in a “super early alpha” state and has drawn criticism, but Thomas promises future fidelity.
Cisco announced at WebexOne a broad Webex expansion that adds AI agents to meetings and calls, introduces customer-service automation, and connects Anthropic and OpenAI models. Most new features are planned for early 2027.
OpenAI published 722 math papers from an unreleased AI model, spanning about 20 subfields and including a proof of the quasi-Riemann hypothesis.
Nous Research raised a $90M Series B at a $1.5B valuation and launched Hermes for Businesses, moving its open-source AI agent into the enterprise.
OpenAI is rolling out new GPT-6 models in ChatGPT with Intelligent UI, which lets the chatbot combine text, graphics, buttons, forms, charts, and interactive tools in responses. Paid users get access first, with Free and Go users following on October 8.
Liquid AI has released Open d1, two open-weight multimodal decision models that return calibrated typed answers with zero output tokens. d1-3B and d1-omni-600M are on Hugging Face and target real-time decisions on NVIDIA hardware.
TechRadar reports Huawei and DeepSeek have released open source AI tools aimed at reducing reliance on Nvidia, while asking whether programmers will switch. The report provides no further details.
Anthropic has released Claude Haiku 5.5, the smallest and cheapest model in its lineup, saying it costs about 75% less to run than the version it replaces and is the company's fastest model to date.
OpenAI began rolling out Intelligent UI and GPT-6 to ChatGPT, adding interactive charts, diagrams and buttons. Common Sense Media simultaneously rated ChatGPT for Teens an 'unacceptable risk,' and OpenAI disputed the findings.
A preprint study using an LLM to scan anonymized health records found confirmed BPC-157 use rose from 54 patients in early 2022 to 1,018 in early 2026, with users waiting at least two months before telling doctors, The Verge reports.