open-source software

Nvidia releases Nemotron 3.5 open AI models for agents

Nvidia on Thursday, Oct. 15, added three open-weight models to its Nemotron family. The flagship is Nemotron 3.5 Super VL, a 120-billion-parameter multimodal model for "specialized always-on agents." It comes with two edge models, Nemotron 3.5 Bolt 74B-A9B 24GB and…

You're all caught up

You've seen all the main stories from the past 2 days.

Reflection AI unveils Beam, its first open-weight model

Reflection AI, a two-year-old startup backed by Nvidia, released Beam on Monday, October 5, 2026. Beam is the company's first open-weight AI model. Reflection says it matches leading Chinese open models on advanced reasoning benchmarks while using much less compute.…

Chinese AI models close gap with US rivals, reshaping global race

Alibaba Cloud's Qwen team released QVQ-72B-Preview, an experimental open-weight model that combines visual understanding with chain-of-thought reasoning, marking an early entry in what has since become a crowded field of multimodal AI systems.

DeepSeek open-sources its first vision model for AI agents

DeepSeek has open-sourced the weights of its first multimodal model, V4-Flash-Vision-Exp, making 305 billion parameters available on Hugging Face under the MIT License. The release, which went live on August 31, marks the completion of a rollout that began August…

OpenClaw 2.0 overhauls open-source AI assistant

The open-source AI assistant OpenClaw released version 2.0 on August 30, its largest update to date, built by 933 contributors and composed of over 16,000 pull requests — roughly half of all code submissions in the project's history.

  • 2 mins

Nvidia agrees to buy Hugging Face for $12.9 billion

Nvidia has agreed to acquire Hugging Face, the open-source AI model platform, for $12.9 billion, according to a report by The Information on Wednesday. Separately, Business Insider reported that the two companies have held acquisition talks in recent weeks at…

Anthropic deploys Mythos 5 in security tools, launches $35M open-source fund

Anthropic announced on Friday that it is committing $35 million in Claude credits to a new fund for securing open-source software, while simultaneously deploying its most capable cybersecurity model, Claude Mythos 5, into enterprise security scanning tools for the first…

Linux kernel maintainers say AI-generated patches have left them ‘completely overwhelmed’

The Linux 7.3 kernel's networking subsystem changes were merged into mainline on Thursday, August 20, bringing extended BIG TCP support for UDP tunnels, reduced global lock contention for routing updates, and new wireless drivers. But alongside the technical gains, networking…

DeepReinforce releases Ornith-1.5, claims parity with Claude Opus 4.8

DeepReinforce released Ornith-1.5 on August 19, 2026, an open-weight family of language models spanning three sizes — 397B, 35B, and 9B parameters — that the company says matches or exceeds Anthropic's Claude Opus 4.8 on several coding and reasoning benchmarks.…

Alibaba launches laptop-ready AI model, challenging Meta

Alibaba on Monday released a new AI model designed to run on consumer laptops and opened the weights of its most powerful system, escalating the Chinese tech giant's competition with Meta for dominance in the open-weight artificial intelligence market.

Alibaba’s Qwen AI models hit 3 billion downloads, passing Meta and Google

Alibaba has open-sourced the weights of its Qwen3.8 model series, including its flagship 2.4-trillion-parameter Qwen3.8-Max and the smaller Qwen3.8-27B, as cumulative global downloads of the Qwen model family surpassed 3 billion — making it the most downloaded open-weight AI model…

Chinese open-weight AI model matches frontier power but lacks safety controls, report finds

A new report from AI safety nonprofit SaferAI has found that Z.ai's open-weight GLM-5.2 model is only months behind the world's leading closed-source AI systems in cyber and biological capabilities — while lacking the safety controls those systems rely on…

China weighs curbing overseas access to its top AI models

China's open-weight artificial intelligence models are making rapid inroads globally, as Moonshot AI's Kimi K3 and a wave of competing Chinese systems offer developers and businesses capable alternatives to proprietary American platforms at a fraction of the cost.

Amazon links North Korean group to four open-source software attacks

Amazon Amazon.com, Inc. disclosed Wednesday that a North Korean hacking group carried out a coordinated campaign against open-source software packages, connecting four separate compromises that security researchers had previously treated as isolated incidents.

Moonshot AI launches Kimi K3, largest open-weight AI model

Chinese AI startup Moonshot AI on Wednesday unveiled Kimi K3, a 2.8 trillion-parameter mixture-of-experts model that it calls the largest open-weight language model ever released, escalating the competition between Chinese and American AI labs at a moment when businesses are…

MiniMax plans 2.7 trillion-parameter open-weight AI model, the largest from a Chinese firm

Chinese AI startup MiniMax Group is developing a large language model with 2.7 trillion parameters that it plans to release as open-weight software as early as this quarter, according to a report by Reuters on Tuesday, in what would be…

IBM and Red Hat commercially launch Lightwell open-source security platform

IBM International Business Machines Corporation and Red Hat on Tuesday announced the commercial launch of Lightwell, a platform delivering automated vulnerability remediation at scale for enterprises navigating the risks of open-source software in cloud and AI environments.

Five Eyes warns China’s open-source AI model poses cyber threat

A frontier-class AI model released by Chinese lab Z.ai (formerly Zhipu AI) is raising fresh cybersecurity concerns among intelligence officials and security researchers, who warn that open-source systems nearly as capable as the most advanced commercial models are already being…

Z.ai’s open-weights GLM-5.2 beats GPT-5.5 on coding benchmarks

Chinese AI startup Z.ai, formerly known as Zhipu AI, released GLM-5.2 as an open-weights model on June 16, delivering a 753-billion-parameter system that outperforms OpenAI's GPT-5.5 on several long-horizon coding benchmarks while costing roughly one-sixth as much to operate via…

Yale proposes copyleft rules for AI trained on open-source code

Researchers at Yale's Digital Ethics Center have introduced a novel licensing framework that would require AI developers who train models on open-source code to make their architectures and training data publicly available, extending a decades-old principle from the free software…