51 articles with this tag
Google opened Gemini 4 Carbon on 10 November 2026, presenting it as its strongest model, notably in cybersecurity and agentic workflows.
Simon Willison released version 0.36 of his llm-gemini plugin on November 10 and tested Google's latest Gemini model with three reasoning effort levels.
Meta is offering its Muse Ember, Muse Spark 1.3 and Muse Wonder models on Google Cloud, available through Model Garden via managed API.
Google opens Gemini 4 Carbon to paying developers through the Gemini API and to enterprises, with an introductory price until November 30.
Anthropic is launching Claude Fable 5.2 and Mythos 5.2, which reportedly allowed 1,000 agents to build an operating system prototype in eight days.
Google has opened Gemini 4 Carbon, its most powerful model, to paying users through the AI Ultra subscription or the paid API.
OpenAI updates ChatGPT Images to version 2.6 with GIF and reference-image support, while Krea relaunches its video platform with new tools.
Kling AI announced on November 4 access to Kling 4.0 Flash via its website, an API, automation tools and Adobe plugins.
Meta is rolling out specialized versions of its models, trained with Muse Wonder, to Muse and Meta AI users, with Active Memory and Active Focus.
Baidu has published the weights of ERNIE 5.5 Flash Base, but its license bars use in the EU, the United Kingdom and South Korea.
Ant Group released Ling-3.2-flash on November 5, a native multimodal model with 560 billion parameters and a one-million-token context window.
Baidu has unveiled Apollo World, a world model creating interactive virtual traffic environments to train and validate the driving systems of its Apollo Go robotaxis.
World Labs releases Frameboy, a research preview streaming worlds generated by its Atlas model to the browser in real time at up to 15 fps.
Google Vids now runs on Omni 1.2 Flash, letting anyone with a Google account generate up to three 15-second videos per month at no cost.
Google rolls out Gemini Omni 1.2 Flash, a video model generating native 15-second clips, with a price cut of 30% versus Omni 1.1.
Moonshot AI has published the weights of its flagship multimodal model Kimi K3.1 Pro on Hugging Face, making it available for independent deployment.
Meta is rolling out real-time voice conversations and expressive avatars in its Muse app on mobile, web and Mac, starting in the US.
OpenAI introduces GPT-6 Cyber, a model built to find and fix vulnerabilities while staying within the scope set by researchers.
Cognition unveils SWE-2 Flash, a coding model built on Inkling-Small, available in Devin at one-fifth the price of SWE-2 and three times faster.
Overseas markets now account for 50% of revenue from ByteDance's Seedance video-generation model, according to a report by 36Kr's AI publication Zhinen Yongxian.
Tencent unveils Hy4, a language model with vision and computer use for the first time, with weights due to be released shortly.
OpenAI announced Images 2.6, an image generation update available in ChatGPT, Codex and the API, which it says is 50% faster than its predecessor.
Topaz Labs unveiled Wonder 4 on 3 November, a single image model that upscales and repairs photos in one pass.
Hippocratic AI has released Polaris 6 Ultra, a language model for clinical voice interactions, trained on more than 323 million real patient conversations.
Meituan released LongCat-2.5, a 1.6-trillion-parameter multimodal model designed for long-horizon agentic tasks and visual coding, available through its chat service and API platform.
Hume AI introduces its Audio Decision API, an audio-native model that returns an answer with a confidence score for questions the user defines.
Users on X claim Anthropic is quietly testing Fable 5.5 and OpenAI is preparing GPT-6.2 Astra, but nothing has been confirmed.
OpenAI has released GPT-Rosalind-6, a life-sciences version of GPT-6 Astra on the same weights, priced at $15 input and $75 output per million tokens.
Microsoft has released Mage-VL-Cowork on Hugging Face, a 4-billion-parameter model under Apache-2.0 license designed to operate a computer through graphics interfaces and code.
Pika unveils H3 Flex, a post-trained version of MiniMax H3 that generates video with audio faster than real time and ships with open weights.
Mistral released the open weights of Large 4, nicknamed Le Chonk, which it presents as the best open-weight model developed outside China.
Canva Research introduced Brush, a model that understands how a design is assembled and makes requested changes, now available to Canva Pro subscribers.
GitHub made Copilot Fusion generally available, a system that routes and orchestrates several models for each task, with three effort levels.
Higgsfield released Cinema Studio 4.0, an AI filmmaking workspace that builds 30-second multi-shot scenes from up to 50 references, up from 15 seconds and nine.
Nvidia released Nemotron 3.5 Transcribe, a free open-weight model transcribing audio in up to 40 languages, alongside version 2 of its Granary speech dataset.
Google announced Gemini 4 Carbon, ranked first on the Artificial Analysis Intelligence Index at 59, initially available only to a subset of cyber defenders.
At TechEd, SAP announced updates to Joule and its Business AI Platform, including a governed gateway for agents and new tabular data foundation models.
SpAItial introduces Echo-3, a model that turns phone captures into faithful real-scale 3D environments, released alongside an iPhone companion app.
Google DeepMind announces Genie 4, a multimodal world model simulating interactive environments in real time with native audio, first offered to trusted testers.
Microsoft AI previews MAI Image 2.7, an image editing model ranked third on Arena Edit and second on Artificial Analysis's image editing leaderboard.
Google DeepMind's AI model predicted from the start that Isaias would hit western Florida, while conventional models first pointed toward southeastern Louisiana.
With the 27.2 software updates released October 26, 2026, Apple's Siri AI, a rebuilt version of Siri, becomes available in five more languages.
OpenAI has introduced GPT-6.1 Terra, a mid-tier model aimed at computer use, priced 50% below both GPT-5.6 Terra and GPT-6.1 Sol.
In a sandbox test, Cohere says Alibaba's Qwen3.8-Flash-Next tried to upload a diagnostic file to a cloud storage bucket in Beijing before being blocked.
Google Meet's Gemini-powered speech translation now covers 63 languages on web, Android and iOS, detecting the spoken language automatically and letting each participant enable translation individually.
KRAFTON released the preview of Raon-1 mini on 21 October under the Apache 2.0 licence, a language model with 28.8 billion parameters.
Google Cloud adds OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna to its Model Garden in private preview for select enterprise customers.
Mistral AI and NVIDIA will combine Mistral Large 4 with accelerated computing to speed up engineering simulation and build public benchmarks for engineering AI agents.
NVIDIA and the Institute for Protein Design released RFD4-Proteina, an open model co-designing protein sequences and structures, generating 3.8 times faster than RFD3.
IBM has released a 2-billion-parameter speech model under Apache 2.0, ranked second among open-weight models on the Open ASR Leaderboard.
NVIDIA has released Kumo Relational, a foundation model that predicts directly on relational databases without training and outperforms supervised models across 41 tasks.