Liquid AI LFM3: Six Open-Weight 2B Models That Fit in 1 GB
Liquid AI launched LFM3 — six open-weight ~2B models for text, vision, audio, omni, and decisions. Runs in 1 GB, fine-tunable on one ...
Hugging Face Transformers 5.21: MiniCPM-OCR and Pocket TTS
Transformers 5.21 adds MiniCPM-OCR, a 1.3B document parser that benchmarks above cloud OCR, and Pocket ...
Kimi K3.1 Pro: The Open-Weight Frontier Narrows to 3 Points
Moonshot AI released Kimi K3.1 Pro, a 2.8T open-weight model that narrowed the frontier gap ...
MLflow 3.9: AI Assistant and MemAlign for Agent Developers
MLflow 3.9 ships a Claude Code-powered AI assistant, MemAlign judge optimizer (30-50% accuracy gain), and ...
TorchNeuron: Run PyTorch on AWS Trainium in One Line
Amazon's TorchNeuron lets existing PyTorch code run on Trainium2 and Trainium3 with one line change. ...
Kimi K3.1 Pro: #4 in the World, Open-Weights Nov 4
Moonshot AI's Kimi K3.1 Pro scores 55 on the global intelligence index—beating GPT-6 Astra at half the price. Full model weights release November ...
DeepSeek Open-Sources DeepInfer: The Production Inference Engine
DeepSeek just open-sourced the inference engine actually running its production services. On October 20, DeepInfer ...
Hugging Face Transformers 5.0: What Breaks and What’s New
Hugging Face Transformers 5.0 drops TensorFlow and Flax, adds native multimodal pipelines, and cuts fine-tuning ...
AMD and Intel Unify PyTorch x86 CPU Backend in oneDNN
AMD is folding ZenDNN into Intel’s oneDNN, giving PyTorch one x86 CPU backend for all ...
Pi 1.1: MCP Ships After a Year of “Never” — Upgrade Now
Pi 1.1 ships MCP via Codemode sandbox, tool modifiers, Claude Haiku 5.5, and OSC 7501. ...