The open‑source AI runtime adds GPU memory mapping, concurrent generation, and a smoother ChatGPT CLI restart.
NoteBy OossaPublished by Oossa: 1 min read
Ollama announced version 0.40.17‑rc0. The update adds mmap support for Vulkan GPUs, letting the model use GPU memory more efficiently. It also enables concurrent text generation in the MLX backend and improves how the ChatGPT command‑line interface restores after a crash. The llama.cpp library was upgraded to commit b12257, and a new contributor made their first pull request.
Developers can run larger models faster on compatible GPUs, reducing latency for local AI applications.
Models
The new world‑action model predicts future video and robot actions, letting robots pick up skills from a single human demonstration.
Models
ChatGPT will let users talk with brand‑verified agents for shopping help, starting with select US e‑commerce firms in beta.
Models
Claude Meetings, Whiteboards and Sheets launch in beta, with different paid-plan access. Anthropic also makes Claude Data generally available.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
