NVIDIA AI(@NVIDIAAI)

Follow along with @llm_wizard below or read more here: https://t.co/N7qu5jb2oL

NVIDIA AI (@NVIDIAAI) — Thu, 23 Jul 2026 16:23:40 GMT

Follow along with @llm_wizard below or read more here: developer.nvidia.com/blog/start-cus…

Your browser does not support the video tag.

🔗 View on Twitter

💬0🔄2❤️8👀1118📊2

⚡ Powered by xgo.ing

A hosted RL run took Nemotron 3 Nano from 22% to 91% accuracy on a math task for under $5. Using @P...

NVIDIA AI (@NVIDIAAI) — Thu, 23 Jul 2026 16:04:10 GMT

A hosted RL run took Nemotron 3 Nano from 22% to 91% accuracy on a math task for under $5.

Using @PrimeIntellect Lab, the whole loop runs on hosted infrastructure. You check a baseline, train the model until the reward climbs, and retest to confirm it actually learned, ending with a downloadable LoRA adapter.

You can also scale this workflow to Nemotron 3 Super and Ultra by changing one line.

💬17🔄12❤️123👀8371📊34

⚡ Powered by xgo.ing

Thanks to @googlecloud for helping power the challenge with G4 VMs featuring NVIDIA RTX PRO 6000 Bla...

NVIDIA AI (@NVIDIAAI) — Wed, 22 Jul 2026 19:21:40 GMT

Thanks to @googlecloud for helping power the challenge with G4 VMs featuring NVIDIA RTX PRO 6000 Blackwell GPUs.

💬1🔄1❤️11👀2771📊2

⚡ Powered by xgo.ing

https://t.co/G4nMai12tC

NVIDIA AI (@NVIDIAAI) — Wed, 22 Jul 2026 19:21:40 GMT

x.com/refv00/status/…

re@refv00

Our team won 1st place in the NVIDIA Nemotron Model Reasoning Challenge on Kaggle🏆

Key idea: train the model to memorize minimal problem patterns and their candidate solutions, then search and verify them at inference.

Thanks @NVIDIA@Kaggle!

Writeup:doi.org/10.34740/kaggl…M

🔗 View Quoted Tweet

💬0🔄1❤️5👀4048📊1

⚡ Powered by xgo.ing

Last month, the NVIDIA Nemotron Model Reasoning Challenge wrapped on @kaggle with 5,000+ participant...

NVIDIA AI (@NVIDIAAI) — Wed, 22 Jul 2026 19:21:39 GMT

Last month, the NVIDIA Nemotron Model Reasoning Challenge wrapped on @kaggle with 5,000+ participants across 4,000+ teams.

The community explored techniques for improving reasoning accuracy, here's what we learned 👇

💬16🔄26❤️185👀13108📊40

⚡ Powered by xgo.ing

Build and Optimize Robotics and Edge AI with the New Jetson Thor T2000 and T3000 https://t.co/xbqx5X...

NVIDIA AI (@NVIDIAAI) — Wed, 22 Jul 2026 18:00:57 GMT

Build and Optimize Robotics and Edge AI with the New Jetson Thor T2000 and T3000 x.com/i/broadcasts/1…

💬4🔄12❤️81👀10018📊17

⚡ Powered by xgo.ing

The new 4-step Cosmos 3 Super models generate images and video up to 25x faster than the originals, ...

NVIDIA AI (@NVIDIAAI) — Wed, 22 Jul 2026 15:19:08 GMT

The new 4-step Cosmos 3 Super models generate images and video up to 25x faster than the originals, and still rank among the best open-weight models on @ArtificialAnlys.

🥇 #1 for image-to-video (no audio)
🥈 #2 for text-to-image

Try them on @huggingfacevist.ly/5c4kesK

💬23🔄84❤️539👀40210📊121

⚡ Powered by xgo.ing

Congratulations to the students who competed at the International Mathematical Olympiad (IMO) 2026. ...

NVIDIA AI (@NVIDIAAI) — Tue, 21 Jul 2026 19:01:27 GMT

Congratulations to the students who competed at the International Mathematical Olympiad (IMO) 2026. 👏

We put Nemotron 3 Ultra to the test to take on the same problems in the same time limit, with no internet or external tools.

The IMO team graded its solutions 30/42, above the 29-point gold threshold. 🥇

💬26🔄85❤️1109👀309801📊164

⚡ Powered by xgo.ing

Devin Outposts are now available on NVIDIA Brev. Write and profile kernels, run experiments, serve ...

NVIDIA AI (@NVIDIAAI) — Tue, 21 Jul 2026 18:10:40 GMT

Devin Outposts are now available on NVIDIA Brev.

Write and profile kernels, run experiments, serve and fine-tune OSS models, and test against real hardware. 

Try it out: brev.nvidia.com/launchable/dep…

Cognition@cognition

Introducing Devin Outposts: run Devin on any machine.

Your Mac mini, a GPU box in your lab, a VM inside your private network, or a Kubernetes cluster next to your internal services.

🔗 View Quoted Tweet

💬11🔄12❤️128👀18407📊23

⚡ Powered by xgo.ing

Tune the Harness, Before Tuning the Model with LangChain | Nemotron Labs https://t.co/0Wja9NSwui

NVIDIA AI (@NVIDIAAI) — Tue, 21 Jul 2026 18:00:00 GMT

Tune the Harness, Before Tuning the Model with LangChain | Nemotron Labs x.com/i/broadcasts/1…

💬6🔄17❤️100👀9809📊24

⚡ Powered by xgo.ing

Congrats to the @poolsideai team on their latest release, Laguna S 2.1. It’s open-weight, delivers ...

NVIDIA AI (@NVIDIAAI) — Tue, 21 Jul 2026 17:08:07 GMT

Congrats to the @poolsideai team on their latest release, Laguna S 2.1.

It’s open-weight, delivers way beyond its size, runs great locally and you can customize it with NVIDIA NeMo. 

Go try it out on @OpenRouter or download from @huggingface �huggingface.co/poolside/Lagun…O

Poolside@poolsideai

Today we're releasing Laguna S 2.1, our most capable model to date.

It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.

Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark.

Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingfacepoolside.ai/blog/introduci…

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬20🔄69❤️670👀57284📊93

⚡ Powered by xgo.ing

NVIDIA Blackwell Ultra just set a world record for pre-training DeepSeek-V3 671B, achieving 1,648 TF...

NVIDIA AI (@NVIDIAAI) — Tue, 21 Jul 2026 15:03:04 GMT

NVIDIA Blackwell Ultra just set a world record for pre-training DeepSeek-V3 671B, achieving 1,648 TFLOPs per GPU.

This is about 3x the earlier delivered performance of the previous generation, and is a result of our extreme co-design and continuous software optimization across popular frameworks including Megatron-Core, TorchTitan, and JAX.

💬42🔄120❤️1329👀108751📊200

⚡ Powered by xgo.ing

👀

NVIDIA AI (@NVIDIAAI) — Mon, 20 Jul 2026 19:35:16 GMT

👀

Poolside@poolsideai

Turns out, the next Laguna also fits on a @NVIDIAAI DGX Spark 👀

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬22🔄22❤️318👀30629📊44

⚡ Powered by xgo.ing

As always, Cosmos 3 Edge is fully open. This includes model weights, post-training recipes and code....

NVIDIA AI (@NVIDIAAI) — Mon, 20 Jul 2026 16:05:19 GMT

As always, Cosmos 3 Edge is fully open. This includes model weights, post-training recipes and code.

Available now on @huggingface 🤗 
nvda.ws/4bMtY59y

💬8🔄54❤️328👀31211📊64

⚡ Powered by xgo.ing

Among open models of a similar size, Cosmos 3 Edge ranks #1 on VANTAGE-Bench for vision analytics.

NVIDIA AI (@NVIDIAAI) — Mon, 20 Jul 2026 16:05:18 GMT

Among open models of a similar size, Cosmos 3 Edge ranks #1 on VANTAGE-Bench for vision analytics.

💬2🔄2❤️53👀7839📊8

⚡ Powered by xgo.ing

Introducing Cosmos 3 Edge: our open frontier world model built to run on-device. Cosmos 3 Edge help...

NVIDIA AI (@NVIDIAAI) — Mon, 20 Jul 2026 16:05:16 GMT

Introducing Cosmos 3 Edge: our open frontier world model built to run on-device.

Cosmos 3 Edge helps robots learn and act, autonomous vehicles understand road scenes and predict intent, and vision AI agents reason across live video for smart infrastructure.

With 4B parameters and a 2B Nemotron-based reasoner, you can run it on DGX Spark, NVIDIA Jetson, and more.

Your browser does not support the video tag.

🔗 View on Twitter

💬70🔄242❤️2152👀201732📊440

⚡ Powered by xgo.ing

@lucasfulks Nice rack

NVIDIA AI (@NVIDIAAI) — Sat, 18 Jul 2026 13:28:47 GMT

@lucasfulks Nice rack

💬4🔄2❤️34👀4523📊7

⚡ Powered by xgo.ing

Another leaderboard win for Nemotron 3 Embed 👀 The 8B model takes #1 on LMEB, with the 1B model ri...

NVIDIA AI (@NVIDIAAI) — Fri, 17 Jul 2026 19:46:32 GMT

Another leaderboard win for Nemotron 3 Embed 👀

The 8B model takes #1 on LMEB, with the 1B model right behind it at #2.

LMEB tests whether embedding models can find the right details across long-running conversations and other memory-heavy tasks, useful for agents that need to remember context across sessions.

NVIDIA AI@NVIDIAAI

Today we released Nemotron 3 Embed 8B and it reached #1 overall on RTEB 🏆

RTEB benchmarks retrieval accuracy across real-world tasks. Better retrieval gives agents more relevant context, helping improve response accuracy.

🔗 View Quoted Tweet

💬24🔄52❤️316👀30969📊56

⚡ Powered by xgo.ing

NeMo AutoModel now supports @huggingface Diffusers 🤗 AutoModel is our open source fine-tuning libr...

NVIDIA AI (@NVIDIAAI) — Fri, 17 Jul 2026 16:04:06 GMT

NeMo AutoModel now supports @huggingface Diffusers 🤗

AutoModel is our open source fine-tuning library. It first tackled Transformers, and now supports image and video models too, with ready-to-run full and LoRA recipes so you spend less time wiring up distributed training.

💬20🔄40❤️193👀30561📊49

⚡ Powered by xgo.ing

@huggingface See what’s new 👇 https://t.co/jy1hf4bidv

NVIDIA AI (@NVIDIAAI) — Fri, 17 Jul 2026 16:04:06 GMT

@huggingface See what’s new 👇nvda.ws/44BW3Z5v

💬0🔄2❤️14👀4220📊3

⚡ Powered by xgo.ing

@thinkymachines @huggingface We just added three ways to try out the new @thinkymachines Inkling mod...

NVIDIA AI (@NVIDIAAI) — Thu, 16 Jul 2026 16:46:24 GMT

@thinkymachines@huggingface We just added three ways to try out the new @thinkymachines Inkling model:

Free GPU-accelerated endpoints ➡️ build.nvidia.com/thinkingmachin…
Download and run it with NVIDIA NIM ➡️ catalog.ngc.nvidia.com/orgs/nim/think…
Deploy it with the NVIDIA Dynamo recipe ➡️ github.com/ai-dynamo/dyna…

💬2🔄2❤️19👀3958📊6

⚡ Powered by xgo.ing

Post-Train NVIDIA Cosmos 3 In a Day with NVIDIA TAO Agent Skills | Cosmos Labs https://t.co/yilZ8nZh...

NVIDIA AI (@NVIDIAAI) — Thu, 16 Jul 2026 16:04:07 GMT

Post-Train NVIDIA Cosmos 3 In a Day with NVIDIA TAO Agent Skills | Cosmos Labs x.com/i/broadcasts/1…

💬8🔄8❤️107👀10059📊23

⚡ Powered by xgo.ing

Explore the models and results below. https://t.co/bWjxKLLoHl

NVIDIA AI (@NVIDIAAI) — Thu, 16 Jul 2026 16:03:00 GMT

Explore the models and results below. huggingface.co/blog/nvidia/ne…

💬1🔄2❤️26👀6135📊6

⚡ Powered by xgo.ing

Today we released Nemotron 3 Embed 8B and it reached #1 overall on RTEB 🏆 RTEB benchmarks retrieva...

NVIDIA AI (@NVIDIAAI) — Thu, 16 Jul 2026 16:02:57 GMT

Today we released Nemotron 3 Embed 8B and it reached #1 overall on RTEB 🏆

RTEB benchmarks retrieval accuracy across real-world tasks. Better retrieval gives agents more relevant context, helping improve response accuracy.

💬40🔄117❤️752👀98120📊171

⚡ Powered by xgo.ing

You can also read more below. https://t.co/nUX1Lp8mHv

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 23:02:31 GMT

You can also read more below.

developer.nvidia.com/blog/build-a-m…

💬1🔄2❤️19👀5868📊5

⚡ Powered by xgo.ing

NVIDIA DeepStream 9.1 is here, with 13 agentic skills for building video analytics pipelines. Inst...

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 23:00:00 GMT

NVIDIA DeepStream 9.1 is here, with 13 agentic skills for building video analytics pipelines. 

Instead of manually building your vision AI pipeline from scratch, describe what you want in plain natural language. Use skills with a coding agent, like Claude Code or Codex, to easily handle setup, configuration, and execution. 

New skills include Multi-View 3D Tracking (MV3DT) for tracking objects across multiple cameras, and AutoMagicCalib for automatically calibrating a camera network. 

This release also brings NVIDIA JetPack 7.2 support for edge deployment on Jetson Orin and Thor. All open source on GitHub, check it out: nvda.ws/4vxQ6Yk

Your browser does not support the video tag.

🔗 View on Twitter

💬28🔄87❤️645👀48346📊170

⚡ Powered by xgo.ing

@StockMKTNewz Night 1 ✅

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 22:51:49 GMT

@StockMKTNewz Night 1 ✅

💬5🔄147❤️612👀110479📊108

⚡ Powered by xgo.ing

Congrats @thinkymachines on the new open model 🙌 Inkling was trained on NVIDIA GB300 NVL72 and the...

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 18:15:00 GMT

Congrats @thinkymachines on the new open model 🙌

Inkling was trained on NVIDIA GB300 NVL72 and the NVFP4 checkpoint is available today on @huggingface:huggingface.co/thinkingmachin…3

Happy building!

Thinking Machines@thinkymachines

Today, we are introducing Inkling.

Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.

thinkingmachines.ai/news/introduci…

Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵

🔗 View Quoted Tweet

💬25🔄92❤️1469👀105580📊151

⚡ Powered by xgo.ing

DGX Spark Live: Autonomous AI Agent Migration https://t.co/zvxDmhvPxg

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 17:59:41 GMT

DGX Spark Live: Autonomous AI Agent Migration x.com/i/broadcasts/1…

💬5🔄17❤️88👀10635📊19

⚡ Powered by xgo.ing

and the 🦞group pic. Huge thank you to everyone who joined us!

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 13:52:40 GMT

and the 🦞group pic. Huge thank you to everyone who joined us!

💬11🔄17❤️210👀36332📊33

⚡ Powered by xgo.ing

First stop in Japan: Build-a-Claw. 🇯🇵 Jensen made a surprise stop in the heart of Tokyo to meet w...

NVIDIA AI (@NVIDIAAI) — Wed, 15 Jul 2026 13:11:34 GMT

First stop in Japan: Build-a-Claw. 🇯🇵

Jensen made a surprise stop in the heart of Tokyo to meet with devs, give away some DGX Sparks, and see what they're building with open models.

Your browser does not support the video tag.

🔗 View on Twitter

💬38🔄95❤️824👀89557📊133

⚡ Powered by xgo.ing

How to Run RL Autoresearch with Agent Skills | Nemotron Labs https://t.co/zVHMsORjl5

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 18:00:42 GMT

How to Run RL Autoresearch with Agent Skills | Nemotron Labs x.com/i/broadcasts/1…

💬11🔄43❤️239👀19948📊69

⚡ Powered by xgo.ing

Read the technical walkthrough 👇 https://t.co/IyHjWFmFyY

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 17:00:06 GMT

Read the technical walkthrough 👇nvda.ws/3T3cRplY

💬0🔄1❤️19👀5163📊3

⚡ Powered by xgo.ing

🚦 The zero-shot model misses a visible traffic signal. After post-training Cosmos 3 Nano with LoR...

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 17:00:05 GMT

🚦 The zero-shot model misses a visible traffic signal. 

After post-training Cosmos 3 Nano with LoRA, it correctly identifies the intersection, and overall WTS validation accuracy increases from 54.41% to 87.14%.

📈 With NVIDIA TAO AutoML, accuracy reaches 93.35%. All in under a day.

💬20🔄30❤️297👀39434📊64

⚡ Powered by xgo.ing

If you want to try it out for yourself, check out the video below or follow along here: https://t.co...

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 16:04:45 GMT

If you want to try it out for yourself, check out the video below or follow along here: nvda.ws/4wIdgMz

Your browser does not support the video tag.

🔗 View on Twitter

💬1🔄7❤️101👀14540📊34

⚡ Powered by xgo.ing

We gave a coding agent a goal and a time budget: build a training environment and teach a vision mod...

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 16:03:29 GMT

We gave a coding agent a goal and a time budget: build a training environment and teach a vision model to count colored stars. 

Using autoresearch with NeMo RL, NeMo Gym, and reusable skills, the agent set up, trained and evaluated the model while the researcher steered the work.

Qwen3-VL-2B went from 25% to 96.9% accuracy, and the agent even proposed the next experiment on its own.

💬89🔄249❤️2609👀197703📊660

⚡ Powered by xgo.ing

Proud to support the open source community. Thanks for the Nemotron shoutout @jmorgan! 🙌

NVIDIA AI (@NVIDIAAI) — Tue, 14 Jul 2026 01:14:39 GMT

Proud to support the open source community. Thanks for the Nemotron shoutout @jmorgan! 🙌

ollama@ollama

U.S. open-source models are quickly gaining ground. 

@Nvidia's newest Nemotron Ultra is fast growing on Ollama and unlocking complex, longer running tasks for developers

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬12🔄21❤️219👀31958📊30

⚡ Powered by xgo.ing

As AI models continue to grow in scale and capability, shaping a model matters just as much as its s...

NVIDIA AI (@NVIDIAAI) — Mon, 13 Jul 2026 23:00:01 GMT

As AI models continue to grow in scale and capability, shaping a model matters just as much as its size. 

We're introducing a new series on AI Model Co-Design exploring the synergy between models and hardware. The first post focuses on how model dimensions influence GPU performance, and how the right design choices improve both system throughput and per-user responsiveness. 

You can read it here: nvda.ws/452Idiy

💬50🔄162❤️1355👀358665📊480

⚡ Powered by xgo.ing

And huge thank you @swyx and AIEWF for providing the platform!

NVIDIA AI (@NVIDIAAI) — Sun, 12 Jul 2026 00:51:02 GMT

And huge thank you @swyx and AIEWF for providing the platform!

💬1🔄1❤️23👀7065📊3

⚡ Powered by xgo.ing

Epic panel - let's run it back soon.

NVIDIA AI (@NVIDIAAI) — Sat, 11 Jul 2026 21:29:12 GMT

Epic panel - let's run it back soon.

Ahmad@TheAhmadOsman

DROP EVERYTHING

The first panel from the Local AI Summit at AIEWF is now live

“State of the Union: Why Local, Why Now”

Featuring leaders from NVIDIA, Roboflow, Osmantic, Forward Future, and EXO Labs

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬9🔄17❤️133👀28217📊24

⚡ Powered by xgo.ing

https://t.co/DRS9kp0hT3

NVIDIA AI (@NVIDIAAI) — Thu, 09 Jul 2026 21:24:15 GMT

x.com/ArashVahdat/st…

Arash Vahdat@ArashVahdat

Heading to #ICML2026 today? Check out our new Flex Forcing that allows autoregressive and bidirectional video generation in a single unified and flexible framework.

🔗 View Quoted Tweet

💬0🔄2❤️16👀6459📊2

⚡ Powered by xgo.ing

Our research team just released Flex-Forcing: a video generation method that lets a single model swi...

NVIDIA AI (@NVIDIAAI) — Thu, 09 Jul 2026 21:24:09 GMT

Our research team just released Flex-Forcing: a video generation method that lets a single model switch between generation methods at inference time. 

Right now there are two main approaches to video generation. Bidirectional diffusion models attend to every frame at once, holding structure well at the cost of speed. Autoregressive models generate frame by frame, so they stream fast and scale to long clips, but accumulate error and drift over time. 

Flex-Forcing trains a single model to do both, letting you choose from the range at inference based on your compute budget.

Your browser does not support the video tag.

🔗 View on Twitter

💬7🔄34❤️225👀31957📊47

⚡ Powered by xgo.ing

Flex-Forcing was also recognized with a spotlight at #ICML2026. You can find the full project page h...

NVIDIA AI (@NVIDIAAI) — Thu, 09 Jul 2026 21:24:09 GMT

Flex-Forcing was also recognized with a spotlight at #ICML2026. You can find the full project page here: nvda.ws/4yw3sqP

💬3🔄3❤️20👀4828📊5

⚡ Powered by xgo.ing

@AtlasInference Lets gooo!

NVIDIA AI (@NVIDIAAI) — Thu, 09 Jul 2026 05:38:05 GMT

@AtlasInference Lets gooo!

💬0🔄1❤️5👀800📊1

⚡ Powered by xgo.ing

@LangChain Here’s a technical walk-through for creating a LangChain Deep Agents harness profile for ...

NVIDIA AI (@NVIDIAAI) — Wed, 08 Jul 2026 16:45:11 GMT

@LangChain Here’s a technical walk-through for creating a LangChain Deep Agents harness profile for Nemotron 3 Ultra.

developer.nvidia.com/blog/create-a-…

💬3🔄3❤️24👀3799📊7

⚡ Powered by xgo.ing

Excited to see this out in the open. We've been working with @LangChain to tune their Deep Agents h...

NVIDIA AI (@NVIDIAAI) — Wed, 08 Jul 2026 16:44:51 GMT

Excited to see this out in the open.

We've been working with @LangChain to tune their Deep Agents harness for Nemotron 3 Ultra. 

The result is a fully open, customizable agent that delivers frontier performance at a fraction of the cost of leading closed models.

LangChain@LangChain

We tuned the harness for @NVIDIAAI Nemotron 3 Ultra.

Benchmark-leading performance. 10x lower inference costs.

✅ An aggregate score of 0.86 at a cost of $4.48
✅ The closest-performing model: $43.48

langchain.com/blog/langchain…

🔗 View Quoted Tweet

💬26🔄57❤️435👀46003📊74

⚡ Powered by xgo.ing

You're welcome 👊

NVIDIA AI (@NVIDIAAI) — Wed, 08 Jul 2026 01:37:24 GMT

You're welcome 👊

mr-r0b0t@mr_r0b0t

@NVIDIAAI just gifted us a 75B MoE 🤩🤩🤩

nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4

75.3B total / 9.3B active compressed from Nemotron-3-Super-120B using the Iterative Puzzle framework.
1M token context support!

Perfect for your single GB10 ♥️

🔗 View Quoted Tweet

💬25🔄33❤️721👀80143📊95

⚡ Powered by xgo.ing

Check out the project page here. https://t.co/8ZzNiVCLc5

NVIDIA AI (@NVIDIAAI) — Tue, 07 Jul 2026 19:28:30 GMT

Check out the project page here. nvda.ws/4pabAZI

💬3🔄2❤️10👀4451📊4

⚡ Powered by xgo.ing

When you train a video model for motion, only some clips actually matter. New from NVIDIA Research ...

NVIDIA AI (@NVIDIAAI) — Tue, 07 Jul 2026 19:28:29 GMT

When you train a video model for motion, only some clips actually matter.

New from NVIDIA Research and recipient of an Outstanding Paper Honorable Mention at #ICML2026: MOTIVE shows exactly which training clips improve motion.

MOTIVE re-weights training signals toward moving regions and away from static backgrounds, then scores each clip by its influence on motion. You can then curate a small high-influence subset to fine-tune on and get better temporal dynamics.

Because it attributes motion rather than appearance, MOTIVE curates data that boosts VBench dynamics and wins 74.1% human preference against the base model.

Your browser does not support the video tag.

🔗 View on Twitter

💬12🔄21❤️198👀17373📊39

NVIDIA AI(@NVIDIAAI)

Follow along with @llm_wizard below or read more here: https://t.co/N7qu5jb2oL

A hosted RL run took Nemotron 3 Nano from 22% to 91% accuracy on a math task for under $5. Using @P...

Thanks to @googlecloud for helping power the challenge with G4 VMs featuring NVIDIA RTX PRO 6000 Bla...

https://t.co/G4nMai12tC

Last month, the NVIDIA Nemotron Model Reasoning Challenge wrapped on @kaggle with 5,000+ participant...

Build and Optimize Robotics and Edge AI with the New Jetson Thor T2000 and T3000 https://t.co/xbqx5X...

The new 4-step Cosmos 3 Super models generate images and video up to 25x faster than the originals, ...

Congratulations to the students who competed at the International Mathematical Olympiad (IMO) 2026. ...

Devin Outposts are now available on NVIDIA Brev. Write and profile kernels, run experiments, serve ...

Tune the Harness, Before Tuning the Model with LangChain | Nemotron Labs https://t.co/0Wja9NSwui

Congrats to the @poolsideai team on their latest release, Laguna S 2.1. It’s open-weight, delivers ...

NVIDIA Blackwell Ultra just set a world record for pre-training DeepSeek-V3 671B, achieving 1,648 TF...

Read more 👇 https://t.co/BXYEe4xhd2

👀

As always, Cosmos 3 Edge is fully open. This includes model weights, post-training recipes and code....

Among open models of a similar size, Cosmos 3 Edge ranks #1 on VANTAGE-Bench for vision analytics.

Introducing Cosmos 3 Edge: our open frontier world model built to run on-device. Cosmos 3 Edge help...

@lucasfulks Nice rack

Another leaderboard win for Nemotron 3 Embed 👀 The 8B model takes #1 on LMEB, with the 1B model ri...

NeMo AutoModel now supports @huggingface Diffusers 🤗 AutoModel is our open source fine-tuning libr...

@huggingface See what’s new 👇 https://t.co/jy1hf4bidv

@thinkymachines @huggingface We just added three ways to try out the new @thinkymachines Inkling mod...

Post-Train NVIDIA Cosmos 3 In a Day with NVIDIA TAO Agent Skills | Cosmos Labs https://t.co/yilZ8nZh...

Explore the models and results below. https://t.co/bWjxKLLoHl

Today we released Nemotron 3 Embed 8B and it reached #1 overall on RTEB 🏆 RTEB benchmarks retrieva...

You can also read more below. https://t.co/nUX1Lp8mHv

NVIDIA DeepStream 9.1 is here, with 13 agentic skills for building video analytics pipelines. Inst...

@StockMKTNewz Night 1 ✅

Congrats @thinkymachines on the new open model 🙌 Inkling was trained on NVIDIA GB300 NVL72 and the...

DGX Spark Live: Autonomous AI Agent Migration https://t.co/zvxDmhvPxg

and the 🦞group pic. Huge thank you to everyone who joined us!

First stop in Japan: Build-a-Claw. 🇯🇵 Jensen made a surprise stop in the heart of Tokyo to meet w...

How to Run RL Autoresearch with Agent Skills | Nemotron Labs https://t.co/zVHMsORjl5

Read the technical walkthrough 👇 https://t.co/IyHjWFmFyY

🚦 The zero-shot model misses a visible traffic signal. After post-training Cosmos 3 Nano with LoR...

If you want to try it out for yourself, check out the video below or follow along here: https://t.co...

We gave a coding agent a goal and a time budget: build a training environment and teach a vision mod...

Proud to support the open source community. Thanks for the Nemotron shoutout @jmorgan! 🙌

As AI models continue to grow in scale and capability, shaping a model matters just as much as its s...

And huge thank you @swyx and AIEWF for providing the platform!

Epic panel - let's run it back soon.

https://t.co/DRS9kp0hT3

Our research team just released Flex-Forcing: a video generation method that lets a single model swi...

Flex-Forcing was also recognized with a spotlight at #ICML2026. You can find the full project page h...

@AtlasInference Lets gooo!

@LangChain Here’s a technical walk-through for creating a LangChain Deep Agents harness profile for ...

Excited to see this out in the open. We've been working with @LangChain to tune their Deep Agents h...

You're welcome 👊

Check out the project page here. https://t.co/8ZzNiVCLc5

When you train a video model for motion, only some clips actually matter. New from NVIDIA Research ...