Introducing NVIDIA Cosmos-3-Nano on Reactor. Two new models, available now via the Reactor API. • Sim-Bimanual predicts what a robot sees as you move its arms and grippers. • Sim-Transfer turns simulations into photorealistic video. Try them now: reactor.inc/cosmos-bimanual

using frontier AI for its intended purpose: going back to hong kong in deus ex (2000). an infinitely explorable version of hong kong which streams in realtime using h3-omni + a jev harness.

Introducing FastH3-Omni. The fastest and highest-quality open-source version of realtime H3 by @MiniMax_AI, post-trained by FastVideo @haoailab with @nuvalab. Exclusive preview on the Reactor API until 10.20.2026. • First and last frame conditioning • Audio input + strong dialogue • Up to 3 reference images • Frontier-level performance

Join our CTO @_bschmidtchen and @nebius for a live webinar on October 15 at 9AM PT: “What Makes Real-Time AI Video Possible?” We’ll cover our custom inference stack, what developers are building on Reactor, and how Nebius keeps interactive sessions fast and dependable under load. Register: nebius.com/events/webinar-sc…

Real time character control demo using @reactorworld

Introducing S2-Editing, a realtime video-to-video model by @ViduAI_official. Transform any video, any way you like. Swap characters, remix styles, and try on new outfits and hairstyles, all without missing a beat. Available Day 0 on Reactor.

Meet S2-Avatar by @ViduAI_official. For the first time, create truly interactive AI personas which you can prompt with audio, text or reference images. Available day 0 on Reactor.

🚀 FastH3 on Reactor just got an upgrade, thanks to FastVideo @haoailab. Stream 720p with better audio and dialogue, sharper detail in high-motion scenes, and more consistent faces from frame to frame, all at the same speed. Get started now: reactor.inc/fast-h3-demo-app

AI models are upgrading at a new speed. OpenAI’s GPT-6 Astra model, released last week, follows a self-controlled “chain-of-thought” reasoning pattern that can evade certain kinds of monitoring. The next value differentiator for improved models may be less about performance and more about auditability. More on Lightwork this week: → ByteDance is working towards releasing a new real-time world model, pressuring Western frontier labs to keep pace. → Personal agents like Meta’s Muse and Instinct are gaining popularity, raising a conversation over personal autonomy with agents. → Some applications are throttling agent traffic. What does this mean for the companies building them? @clairejyz @Machiz @buckymoore 0:00 Welcome, Bucky Moore! 1:12 OpenAI's Astra 3:50 The New AI Model’s Architecture 4:40 AI Safety and Guardrails 8:01 Trusting AI Agents 11:09 The Future of AI Models 12:06 The AI Agent Land Grab 17:09 AI, Trust, and Personal Data 20:12 Improving AI Agent Capabilities 22:11 Enterprise AI Assistants and Personal Context 24:05 AI Agents and the Future of the Web 25:50 ByteDance and World Models 28:19 World Models and Consumer Content 30:56 Reactor and World Model Infrastructure 34:13 Real-Time Video Models and Scaling

I made it as easy as possible to build your own Twitch livestream using H3 Reference Turbo Realtime! Consistent Character references, ever-evolving LLM show-runners, and a one-click deploy on Render! Link to get started in the comments

🚀 Introducing H3 Reference Turbo Realtime, the latest model by @MiniMax_AI. Stream video with audio and guide generation with up to 9 reference images, giving you more control than ever. Available today on Reactor. Start building now!

We are partnering with @viskoai to host a hackathon in San Francisco this Saturday to celebrate the launch of Orbis 1.0. Sign up here: luma.com/gh4256ju

We're proud to partner with @nvidia's SANA team to ship FastH3 with Sol. For the first time, generate clips 3x faster than realtime. Available day 0 via the Reactor API. Try it here: reactor.inc/sandbox?model=fa…