Tag with "Hugging Face"

LiquidAI has launched the 'Liquid Nanos' series, a collection of lightweight AI models optimized for edge computing devices like Raspberry Pi. The models, available in 350M and 1.2B parameter versions, support five key applications: translation, extraction, RAG, tool calling, and mathematical reasoning. Twelve task-specific models are now live on Hugging Face.

Hugging Face has launched its Reachy Mini desktop robot, designed to empower open-source AI developers. Available in two versions, the robot kits aim to simplify building and testing AI applications. The release underscores Hugging Face's commitment to open-source hardware and community-driven innovation.

Hugging Face has unveiled SmolLM3, a 3B-parameter open-source language model that competes with larger 4B models in performance. Featuring dual-mode inference, 128K context support, and multilingual capabilities, it offers efficiency for edge devices while maintaining transparency through full open-source release.

Hugging Face has unveiled SmolLM3, a 3B-parameter open-source model with 128K context support and dual reasoning modes. Despite its compact size, it outperforms competitors like Llama-3.2-3B and supports multiple languages. The release includes full architectural transparency to foster community development.

French AI lab Mistral has released its Magistral series of inference models, featuring enhanced logical reasoning capabilities for mathematics and physics applications. The Small version is now available on Hugging Face, while the Medium version remains in preview.

Hugging Face's open-source LeRobot project is transforming AI robotics development by offering unified hardware interfaces, pre-trained models, and collaborative tools, significantly lowering barriers for researchers and developers.

Apple's MLX-LM framework integrates with Hugging Face, enabling seamless local execution of over 4,400 LLMs on M-series chips without cloud dependency or conversion delays.

IBM and Hugging Face have introduced SmolDocling, a lightweight open-source vision-language model designed to convert complex documents into structured data efficiently. With its innovative DocTags format and compact size, SmolDocling outperforms larger models in tasks like OCR, formula transcription, and code recognition.

1. Hugging Face launched SmolLM2, a set of compact language models designed to be efficient and powerful. 2. The models come in three sizes: 135M, 360M, and 1.7B parameters. 3. SmolLM2-1B outperforms Meta’s Llama1B on key benchmarks, especially in scientific reasoning and common sense. 4. These models are perfect for smartphones and edge devices, not just large data centers. 5. SmolLM2 signals a shift toward efficient AI that’s accessible to more users and companies.