2/ What changed: old Cosmos split the work across separate models — one to understand a scene, one to generate video, one for controlled simulation.
Cosmos 3 fuses everything into a single Mixture-of-Transformers with two towers:
→ a reasoner (the VLM "brain")
→ a diffusion
MACHINE LEARNING
-
Cosmos 3 merges reasoner and diffusion towers in one model
By
–
-
NVIDIA open-sources Cosmos 3, first open omnimodel for physical AI
By
–
1/ NVIDIA just open-sourced Cosmos 3 at GTC Taipei!
— Chubby♨️ (@kimmonismus) 1 juin 2026
It's the first fully open "omnimodel" for physical AI – one model that understands the real world, predicts what happens next, and generates the actions a robot should take.
Weights, code, datasets. All open. And this is… https://t.co/5Y8BbXUWqJ pic.twitter.com/3bOMlwO0B21/ NVIDIA just open-sourced Cosmos 3 at GTC Taipei! It's the first fully open "omnimodel" for physical AI – one model that understands the real world, predicts what happens next, and generates the actions a robot should take. Weights, code, datasets. All open. And this is
-
GrepSeek: Training Search Agents for Direct Corpus Interaction
By
–
GrepSeek Training Search Agents for Direct Corpus Interaction
-
NVIDIA announces Nemotron 3 Ultra, 550B open-weight model
By
–
NVIDIA announced an upcoming release of Nemotron 3 Ultra later this week, a 550B-parameter open-weight model.
— 🚨 AI News | TestingCatalog (@testingcatalog) 1 juin 2026
According to Artificial Analysis, it is positioned as the most intelligent open-weight model from the US lab.
Soon 👀 https://t.co/rSWH9n9oIh pic.twitter.com/8XiWwDRwJzNVIDIA announced an upcoming release of Nemotron 3 Ultra later this week, a 550B-parameter open-weight model. According to Artificial Analysis, it is positioned as the most intelligent open-weight model from the US lab. Soon
-
AI agents bottleneck accelerated 208x by classical operators
By
–
The thing slowing down every AI agent just got 208× faster.
— AlphaSignal AI (@AlphaSignalAI) 1 juin 2026
For years, speeding up AI meant making the language model itself run faster.
But agents changed the bottleneck.
They lean on older tools like clustering and search to verify their work.
These classical operators… pic.twitter.com/tr9R5OdTsGThe thing slowing down every AI agent just got 208× faster. For years, speeding up AI meant making the language model itself run faster. But agents changed the bottleneck. They lean on older tools like clustering and search to verify their work. These classical operators
-
VAST AI’s living world persists whether watched or not
By
–
This isn't a video generator.
— AI Highlight (@AIHighlight) 1 juin 2026
It's a living world that keeps running whether you're watching or not.
VAST AI just changed what "world model" means. https://t.co/MZbVveSEpmThis isn't a video generator. It's a living world that keeps running whether you're watching or not. VAST AI just changed what "world model" means.
-
Model with 1B parameters surprises with agentic capabilities
By
–
We're trending on @huggingface
! Tbh, we undersold this model. It's a lot more capable at agentic tasks than I expected. I keep discovering new capabilities every day, it's crazy for 1B active parameters. -
Ultra-realistic photo of a glamorous woman in a crowded stadium
By
–
Hecho con Seedance 2.0 en Pollo AI
— Nico (@nicos_ai) 1 juin 2026
Prompt:
"Foto ultrarrealista de retransmisión deportiva de una mujer glamurosa sentada entre el
público de un estadio de fútbol abarrotado durante un partido nocturno, con un top de satén sin
mangas de cuello alto en marrón oscuro y pendientes… pic.twitter.com/yR2fWosLGMMade with Seedance 2.0 on Pollo AI Prompt:
"Ultra-realistic sports broadcast photo of a glamorous woman sitting among the crowd of a packed football stadium during a night match, wearing a dark brown satin sleeveless turtleneck top and curls" -
Hermes Agent Now Natively Supported on Windows for Agentic AI Workflows
By
–
WINDOWS USERS, REJOICE
— Charly Wargnier (@DataChaz) 1 juin 2026
If you've been waiting for agentic AI workflows, Hermes Agent is now natively supported on @Windows 🔥
Check out the guide from @NOUS here:
→ https://t.co/5qCZWgTlG1pic.twitter.com/fCB4UsNSF9WINDOWS USERS, REJOICE If you've been waiting for agentic AI workflows, Hermes Agent is now natively supported on @Windows Check out the guide from @NOUS here: → https://
hermes-agent.nousresearch.com/docs/user-guid
e/windows-native
… -
NVIDIA accelerates bounding box detection 10x by removing mandatory token prediction
By
–
🚨 NVIDIA just pulled off something crazy: making bounding box detection 10x faster by ripping out the exact step the entire industry assumed was mandatory ↓
— Charly Wargnier (@DataChaz) 1 juin 2026
Every VLM grounding model treats boxes like sentences, predicting them token by token. It’s inherently slow.
Enter… pic.twitter.com/OE7fxZFF4VNVIDIA just pulled off something crazy: making bounding box detection 10x faster by ripping out the exact step the entire industry assumed was mandatory ↓ Every VLM grounding model treats boxes like sentences, predicting them token by token. It’s inherently slow. Enter