Alibaba says its Qwen open-weight models surpassed 3 billion global downloads in 6 months, overtaking Meta and Google while building a rapidly expanding developer ecosystem.
NVIDIA is expanding its Nemotron family with lightweight agent models, streaming speech recognition, routing software and an open platform for local AI development.
Meta released Muse Glimmer, a 30-billion-parameter open-weight model that runs on a single consumer GPU, bringing customizable agentic AI to personal computers.
Thinking Machines Lab, founded by ex-OpenAI CTO Mira Murati, released its first model Inkling, a 975-billion-parameter open-weight system trained on video.
Hugging Face CEO Clem Delangue tells TechCrunch companies are ditching frontier APIs for in-house open-source models, calling the current frenzy an 'LLM.
Google DeepMind released DiffusionGemma, a parallel text generation model NVIDIA optimized for up to 4x faster performance on local GPUs hardware.
OpenAI's Codex CLI bans goblins and pigeons after GPT-5.1's 'Nerdy' personality flooded later models with creature references.
NVIDIA releases open-source omni-modal model that handles documents, audio, and video in a single forward pass, cutting agent latency by up to 9×.
Canonical will roll out optional AI features in Ubuntu 2026 releases, but many users want an AI-free version or universal kill switch.
Google's Gemma 4 brings frontier multimodal AI to devices with Apache 2.0 licensing, enabling offline AI across phones, laptops, and servers.
Over 512K lines of Claude Code CLI source leaked via NPM source maps, exposing a Tamagotchi-style pet and always-on agent, then Anthropic accidentally tried to scrub thousands of GitHub repos before walking it back.
Bluesky launches Attie, an AI assistant that creates custom social feeds and will eventually let users vibe-code entire apps on the AT Protocol.
