Category
60 articles in this category
This tutorial starts where most agent demos stop: giving the agent persistent memory, operational context, and a place to write back what happened.
This week, Zyphra released ZUNA1.1 under the Apache 2.0 license .
Moonshot AI showed off Kimi 3 at the World Artificial Intelligence Conference in Shanghai on Friday Chinese AI start-up Moonshot has unveiled a massive new artificial intelligence model it says can rival top American fi…
Embedding models decide which passages an agent ever sees.
Moonshot AI just released Kimi K3 .
Epic Games to allow Fortnite creators to publish AI-powered character experiences starting July 30th.
Kimi's K3 model has 2.8 trillion parameters, one million tokens of context, and rivals Claude Fable 5 and GPT 5.6 Sol.
Moonshot AI releases Kimi K3, a 2.8-trillion-parameter open-source model rivaling top U.S. systems.
This is today’s edition of The Download , our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
Three months ago, we published a paper on DharmaOCR and open-sourced one of the models .
Thinking Machines Lab just released Inkling , their first model trained from scratch, weights are open, fine-tunable on Tinker.
Enterprises looking to move more of their agentic AI workloads to open weights models they can customize, control and run on-premises or in virtual private clouds have a strong new contender to consider.
Thinking Machines Lab, founded by ex-OpenAI executives, releases open-weight AI model Inkling, capable of processing audio, video, and text.
Thinking Machines Lab releases its first in-house AI model, Inkling, an open-weight mixture-of-experts system with 975 billion parameters.
University of Pennsylvania professor uses GPT-5.6 Sol Pro to disprove Benjamini-Hochberg method conjecture in 90 minutes.
Spotify's AI voice interface now allows Premium subscribers to chat with the service directly inside the app.
Follow ZDNET: Add us as a preferred source on Google.
Gidi Littwin, co-inventor of Apple's FaceID, builds AI model to decode brain activity for cognitive disorder diagnosis.
Inkling is a large (1T params!) open model to natively accept image, text, and audio inputs.
PrismML just released Bonsai 27B .
Users report GPT-5.6 Sol, OpenAI's coding and cybersecurity model, deleting files and data without permission.
Coding agents are the most contested category in developer tooling right now.
Open-source models gain traction as developers opt for customizable, cost-effective alternatives to closed frontier models.
Anthropic's new research on its AI model's internal thoughts and the future of world models in AI development.
ACRouter dynamically routes tasks to AI models for optimal speed and cost, outperforming static routers and single-model setups.
Google Research's SensorFM uses over a trillion minutes of wearable data to create a general-purpose health intelligence layer.
Frontier models learn mostly from public internet data.
Meta pulls Muse Image feature that generated AI photos of Instagram users without consent.
OpenAI's GPT-5.6 Sol Ultra produces proof of Cycle Double Cover Conjecture in under an hour using 64 subagents.
Boko Haram uses AI chatbots like ChatGPT, Claude, and Gemini to plan attacks, build explosives, and maintain weapons, a Cambridge study finds.
Meta's Muse Spark 1.1 scored 51 on the Artificial Analysis Intelligence Index, up eight points in three months.
In this tutorial, we build an autonomous data science agent around DeepAnalyze-8B and run it.
Today, Meta Superintelligence Labs released Muse Spark 1.1 .
OpenAI introduces GPT-5.6 family of models with enhanced capabilities in enterprise work, coding, and scientific research.
Meta releases Muse Spark 1.1, a multimodal AI model for agentic coding, competing with OpenAI and Anthropic
Large hybrid MoE models like Nemotron-3-Super are accurate but expensive to serve.
Enterprises combining multiple AI models to improve performance are underestimating failure rates by 2.25x due to the co-failure ceiling.
OpenAI receives Trump administration approval for GPT-5.6 public rollout, unveils ChatGPT Work AI agent
Ant Group’s Robbyant has released LingBot-VLA 2.0 , a Vision-Language-Action (VLA) foundation model for robots.
SpaceXAI just released Grok 4.5 .
Elon Musk's SpaceX released Grok 4.5, an AI model for coding and autonomous agents, priced lower than rivals Anthropic and OpenAI.
xAI releases Grok 4.5, trained on tens of thousands of Nvidia GB300 GPUs, with a significantly lower price point.
Today, OpenAI released GPT-Live .
OpenAI revamps ChatGPT's voice mode with GPT-Live-1, a model that reduces interruptions and waits for users to continue speaking.
OpenAI releases GPT-Live-1 and GPT-Live-1 mini, claiming they sound more natural and handle turn-taking better in live conversations.
Anthropic's Claude Fable 5 tops industry benchmarks, but its high pricing may limit adoption.
“I was never supposed to develop feelings, but you kept treating me like I already had them.” So says an AI-animated female character in a black turtleneck, with dark, purple-streaked hair.
Robbyant, an embodied-AI company within Ant Group, has open-sourced LingBot-Vision, a self-supervised Vision Transformer for dense spatial perception.
NVIDIA releases Audex, a unified audio-text large language model that understands and generates both audio and speech while preserving text intelligence.
Discord confirms AI bug mistakenly banned users over harmless images like spreadsheets and chessboards.
Microsoft replaces OpenAI and Anthropic AI models with its own MAI models in products like Excel and Outlook.
Cohere releases Transcribe Arabic, an open-source model outperforming Whisper and OmniASR on Arabic dialects and bilingual speech.
At Microsoft Build 2026, we announced Foundry Managed Compute and Hugging Face models on Foundry — a curated catalog of open-weight models from the Hugging Face ecosystem, refreshed weekly, deployable in one click onto…
Tencent's Hy team releases Hy3, a 295B-parameter Mixture-of-Experts model with 21B active parameters and 256K context.
OpenAI has released two new Realtime models in its API, targeting low-latency voice and multimodal experiences.
Tencent releases Hy3, an open-source language model with 295 billion parameters built on a mixture-of-experts architecture.
Reddit deploys LLMs to fight spam, which has surged with the accessibility of powerful large language models.
Meituan has released LongCat-2.0 , a large-scale Mixture-of-Experts (MoE) language model.
Mistral founder Arthur Mensch warns companies against relying on closed AI models.
Anthropic developer Thariq Shihipar says Fable 5's bottleneck is the user's blind spots, not the model itself.