HomeTagsWorld models

World models - Page 15

Stories from World models come first, newest first, then stories from connected topics (Generative AI, AI Video, AI Models, Humanoid Robots, Robotics, Hugging Face), each labelled with where it comes from. This view is not indexed by search engines.

Latest

NVIDIA Releases Nemotron 3.5 Super VL, an Open Multimodal Agent Model

NVIDIA released Nemotron 3.5 Super VL, an open 120B multimodal MoE for agents. Benchmark and partner figures are NVIDIA and partner claims.

MBZUAI’s IFM and Inception42 Release Jais 3 Open Arabic Models and Jais Speech

MBZUAI's IFM and Inception42 released Jais 3, six open Arabic models from 900M to 375B parameters, plus Jais Speech. Licence not stated.

Agility Robotics–Churchill Capital XI S-4 Declared Effective; Vote Set for 2 November

The SEC declared Agility Robotics' Form S-4 effective for its SPAC merger with Churchill Capital XI; vote set for 2 November.

StepFun Publishes Open Weights for Step 5 Preview Under Its Own Licence, Not Apache 2.0

StepFun published Step 5 Preview open weights on Hugging Face. The licence is StepFun Model License 1.0, not Apache 2.0.

Figma Opens Code Layers and Specs Betas and Adds Payload to Make

At Config India, Figma put code layers and Specs into open beta, made Payload the native backend in Figma Make and added 20 MCP partner integrations.

YouTube Music Starts Labelling Songs Made Mainly With AI

YouTube Music is labelling songs created primarily with generative AI, using partner disclosures and its own detection tools. The main app will follow.

Pentagon Folds Its Autonomous Warfare Group Into a New Task Force Inside SOCOM

The Pentagon's autonomous warfare group is being absorbed into a new task force inside SOCOM, Breaking Defense and Inside Defense report.

Mirelo SFX 2.0 Swaps a Video’s Sound and Keeps the Dialogue

Berlin start-up Mirelo says SFX 2.0 rebuilds a video's effects and ambience for clips up to 10 minutes while keeping speech. fal charges $0.010 a second.

UltraBERT Explained: Open 32K-Context Encoders, and Why “How Much You Hide” Decides What a Model Is Good At

DMIS Lab's UltraBERT is a family of open 32K-context encoders. Explainer: what encoders do, and why the masking finding matters for search and RAG.

Physical Intelligence Opens Pi API Beta for Fine-Tuning Robot Models

Physical Intelligence opened a Pi API beta, by application, letting robot builders fine-tune and run its models on its cloud. No pricing was given.

Latest on TSN

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.