Nvidia just open-sourced its Audio2Face technology, releasing the underlying models, SDK, and complete training framework. The tool analyzes audio to generate real-time facial expressions and lip-syncing for 3D digital characters. This lets game and virtual avatar developers customize and deploy high-fidelity character animations without building facial rigs from scratch.
Alibaba just introduced Qwen3-Omni, a natively end-to-end multimodal model unifying text, images, audio, and video under a single architecture. It boasts a 211ms latency alongside 30-minute audio comprehension, releasing 30B-scale instruct, thinking, and captioner checkpoints for developers.
H Company just open-sourced Holo1.5, a family of 3B, 7B, and 72B models built specifically for Computer Use agents. The update boosts UI element localization accuracy by over 10% compared to its predecessor and handles complex software interfaces like Photoshop and VSCode. All weights are now accessible on Hugging Face.
Google and DeepMind just open-sourced VaultGemma, a 1B-parameter model trained from scratch with differential privacy. By mathematically injecting noise into training, it ensures the model never memorizes sensitive sequences. While privacy constraints pull its benchmark performance back to roughly GPT-2 levels, it provides the first empirical scaling laws for provably private AI.
Google just launched EmbeddingGemma, a compact 308M parameter open embedding model tailored specifically for on-device AI. It allows developers to run semantic search and local RAG pipelines entirely offline, keeping sensitive data private while sharing a tokenizer with Gemma 3n to save memory. High-accuracy retrieval is finally fitting inside lightweight local hardware.
ByteDance just expanded its open-source UI-TARS agent series, rolling out UI-TARS-2 with upgrades across GUI workflows, coding, and tool use. Built on native vision-language models with reinforcement learning reasoning, it translates screenshots into system clicks, drags, and keystrokes. Why it matters: It turns complex computer and mobile tasks into reliable, hands-free automation.
Tencent just open-sourced HunyuanVideo-Foley, an end-to-end text-and-video-to-audio framework designed to generate realistic Foley sound effects for videos. Trained on 100,000 hours of multimodal data, it uses an MMDiT architecture to tightly align sound with visual motion, outperforming existing open-source baselines in audio quality and temporal alignment.
OpenAI reminded developers that its Codex CLI coding agent is already fully open-source under the Apache-2.0 license. The lightweight tool runs locally in terminal environments across macOS, Linux, and Windows, integrating with IDEs like VS Code and Cursor. Why it matters: users can run local terminal coding agents powered directly by their ChatGPT plans without waiting on closed releases.
xAI just open-sourced Grok 2.5, releasing the weights for its top 2024 model on Hugging Face under a custom community license. It is free for research and non-commercial projects but bans training rival models. Musk also promised Grok 3 will go open source in about six months.
Elon Musk just announced that xAI will open-source its Grok 2 chatbot next week, sticking to his promise of releasing previous-generation models as new ones arrive. While giants like OpenAI lean into proprietary APIs, xAI is betting that handing developers the actual code will carve out a serious competitive edge.
Researchers at The University of Hong Kong just launched OpenCUA, an open source framework designed to train AI agents that control computers across Windows, macOS, and Ubuntu. Powered by 22,600+ screen-recorded task demonstrations and chain-of-thought reasoning, it matches proprietary models from Anthropic and OpenAI. Why it matters: It finally gives the research community an open recipe to study and build general-purpose desktop agents.
ByteDance's Seed Team just open-sourced Seed-OSS-36B on Hugging Face under the Apache-2.0 license. The 36B-parameter model boasts a 512K token context window—doubling GPT-5's limit—and lets developers customize "thinking budgets" to balance reasoning quality with latency. It gives enterprises a fully permissive, commercially viable alternative for heavy coding and long-document reasoning.
Nvidia just dropped Nemotron-Nano-9B-v2, a hybrid Mamba-Transformer model slimmed down to fit on a single A10 GPU. Developers can toggle reasoning traces on or off using simple /think tokens and dial in exact thinking budgets. It delivers up to 6x faster inference than similar transformers while outperforming Qwen3-8B across key benchmarks.
NVIDIA just released Granary, an open-source multilingual speech dataset spanning around a million hours of audio across 25 European languages. Alongside it, the company launched Canary-1b-v2 for high-accuracy translation and Parakeet-tdt-0.6b-v3 for fast, real-time transcription on Hugging Face.
Google just released Gemma 3 270M, an ultra-compact 270-million-parameter open model built specifically for task-specific fine-tuning. Packed with out-of-the-box instruction following and structuring capabilities, it lets developers ditch bloated models in favor of cheap, fast specialized workers for jobs like text classification and data extraction.
