Search results

A new open speech-to-speech pipeline combines Cerebras fast inference, Gemma 4, and modular components for natural, low-latency voice interactions.

Mount Hub repos directly into any cloud job with no egress fees. Use hf:// URLs and HF_TOKEN to run on 20+ clouds.

LeRobot v0.6.0 introduces world model policies that imagine the future, new VLAs, unified reward models, six new benchmarks, and faster data loading.

We assembled PRX's pre-training data from public and internal sources, re-captioned images with a VLM, and built a streamable corpus for training.

NVIDIA's new Nemotron 3 Embed collection, led by an 8B model that ranks #1 on the RTEB leaderboard, delivers state-of-the-art retrieval quality and production-ready deployment options for RAG, agentic retrieval, and code search.

The transformers modeling backend for vLLM now achieves native-level inference speed for many LLM architectures, letting model authors use their existing code without porting.