Google releases SigLIP 2, a new family of multilingual vision-language encoders that improve upon SigLIP with additional training objectives including a decoder for localization, self-distillation…
Hugging Face released SmolVLM2, a family of small video understanding models in three sizes (2.2B, 500M, and 256M parameters), with MLX support and demos including an iPhone app and VLC integration.
Hugging Face announces the addition of three new serverless inference providers (Hyperbolic, Nebius AI Studio, and Novita) to the Hub, integrated into the UI and client SDKs, enabling access to…
The source text is a brief introduction to a conversation with Uber's Head of AI and Product, Jai Malkani, about how Uber uses AI to enable on-demand experiences.