NVIDIA Open-Sources Its Run:ai Platform for Sharing GPUs on Kubernetes
NVIDIA released the full Run:ai platform, which schedules AI jobs across shared GPU clusters, under Apache 2.0. It includes four parts teams can adopt separately.
6 curated artificial intelligence stories from Tuesday, October 20, 2026 — AI model updates, product launches, and research breakthroughs of the day.
NVIDIA Open-Sources Its Run:ai Platform for Sharing GPUs on Kubernetes
NVIDIA released the full Run:ai platform, which schedules AI jobs across shared GPU clusters, under Apache 2.0. It includes four parts teams can adopt separately.
Timestral is a family of four time-series models, 5M to 209M parameters, that predict future values such as sensor readings or demand without retraining. Weights are open under a modified MIT license.
Drex Max is a decision model: it reads text and images, then returns a probability for each option you give it, instead of writing text. It costs $0.06 per million input tokens.
M3.1 Flash is a 428B-parameter model that activates about 23B per token and reads text, images and video. It costs $0.15 per million input tokens, with weights due Oct 23.
Odyssey says a simulated fruit fly brain, built from a full wiring map of about 140,000 neurons, now flies through worlds generated in real time by its Odyssey-3 model.
SKY-001 turns an image or sketch into step-by-step plans for building parts in CAD and BIM software. Skyfall claims it is 51 to 125 times faster and 150 to 210 times cheaper than top models.
Explore daily AI news digests, model launches, and research breakthroughs from other dates.
Perplexity Launches Frontier Research to Build an AI That Does Research
Google Launches Gemini 4 Models and AI Tools in Busy Week
Liquid AI Introduces LiquidCodec for Fast On-Device Voice AI
ChatGPT Finances Reaches the Desktop App for US Users
Alibaba Previews Qwen4-Max, the First Model of Its Qwen4 Series
OpenAI Lets One Dot Work Across Several of Your Computers