NVIDIA AI(@NVIDIAAI)

Another leaderboard win for Nemotron 🏆 Nemotron 3 Embed 8B ranks #1 for combined nDCG@10 on the Q2...

NVIDIA AI (@NVIDIAAI) — Thu, 10 Sep 2026 21:49:57 GMT

Another leaderboard win for Nemotron 🏆

Nemotron 3 Embed 8B ranks #1 for combined nDCG@10 on the Q2D-Web benchmark, tested across 190M web documents and nearly 70K agent-reformulated queries in 10 languages.

Shoutout to @perplexity_ai for putting it together.

💬17🔄18❤️198👀51140📊41

@USC Congrats to the winners of the AI City Challenge and being featured in the #ECCV2026 Daily News...

NVIDIA AI (@NVIDIAAI) — Thu, 10 Sep 2026 19:41:58 GMT

@USC Congrats to the winners of the AI City Challenge and being featured in the #ECCV2026 Daily News. x.com/tangzhengthoma…

Zheng Tang@tangzhengthomas

Great to see NVIDIA #Cosmos featured in #ECCV2026 Daily News. Team Qyn’s Quang Minh Dinh (@SFU, @SFU_CompSci) and Tuan-Kiet Doan (@IP_Paris_) adapted Cosmos 3 Nano for generative traffic video forecasting and won AI City Challenge Track 5 -- Congratulations!
rsipvision.com/ECCV2026-Thurs…

💬1🔄2❤️9👀4486📊2

Most diffusion-based video models are trained on short clips. When they’re used to relight longer vi...

NVIDIA AI (@NVIDIAAI) — Thu, 10 Sep 2026 19:37:12 GMT

Most diffusion-based video models are trained on short clips. When they’re used to relight longer videos in chunks, the lighting can visibly jump between them.

Meet HorizonRelight, new work from our research team and @USC. It propagates target-domain context from one sliding window to the next, helping each section pick up where the last one left off.

This means more consistent lighting across long videos, with fewer boundary artifacts and unwanted appearance changes.

#ECCV2026 paper + demo: nvda.ws/4xDzAYt

Your browser does not support the video tag.

💬6🔄12❤️214👀14496📊38

Multimodal models put different demands on vision encoding, prefill and decoding. Separating visio...

NVIDIA AI (@NVIDIAAI) — Thu, 10 Sep 2026 18:27:28 GMT

Multimodal models put different demands on vision encoding, prefill and decoding. 

Separating vision encoding from the other stages can reduce resource contention and help models respond faster, but only for the right workloads.

See how EPD disaggregation works, when it helps and what to consider before using it: nvda.ws/4xUFUv2

💬12🔄14❤️197👀14198📊34

Seattle DGX Spark Hackathon Winners Spotlight https://t.co/Tvrr02JtWF

NVIDIA AI (@NVIDIAAI) — Thu, 10 Sep 2026 18:01:09 GMT

Seattle DGX Spark Hackathon Winners Spotlight x.com/i/broadcasts/1…

💬6🔄7❤️60👀10680📊13

Commodore 64 🤝 Nemotron 3.5 Lightning

NVIDIA AI (@NVIDIAAI) — Wed, 09 Sep 2026 20:12:11 GMT

Commodore 64 🤝 Nemotron 3.5 Lightning

Duncan Anderson 🍃🔋⚡️🤖@duncsand

Here I am chatting with @NVIDIAAIDev Nemotron 3.5 Lightening from my Commodore 64! Who says you can’t teach an old dog new tricks?

💬10🔄9❤️98👀13792📊16

Looking for reasons to enter the @NVIDIAGTC Berlin Golden Ticket Contest? 🎫 Here are three. Tomorr...

NVIDIA AI (@NVIDIAAI) — Wed, 09 Sep 2026 14:42:18 GMT

Looking for reasons to enter the @NVIDIAGTC Berlin Golden Ticket Contest? 🎫 Here are three.

Tomorrow is your last chance to enter. Share a short video or link to your open model project, tag the judge you heard about the contest from, and include#NVIDIAGTCC for a chance to win.

Contest details:nvda.ws/4ymwgAY3

Your browser does not support the video tag.

💬9🔄11❤️82👀15998📊19

Visual AI needs to recognize a scene, understand what happened and predict what could happen next. ...

NVIDIA AI (@NVIDIAAI) — Tue, 08 Sep 2026 19:39:28 GMT

Visual AI needs to recognize a scene, understand what happened and predict what could happen next.

For the AI City Challenge at #ECCV2026, teams built systems to do just that. Four of the top five video solutions used NVIDIA Cosmos ⬇️ x.com/i/article/2097…

💬23🔄40❤️204👀15897📊54

Ask the Experts: How NVIDIA OpenShell Secures Autonomous Agents | Nemotron Labs https://t.co/avs8j8i...

NVIDIA AI (@NVIDIAAI) — Tue, 08 Sep 2026 16:59:22 GMT

Ask the Experts: How NVIDIA OpenShell Secures Autonomous Agents | Nemotron Labs x.com/i/broadcasts/1…

💬7🔄6❤️78👀11582📊16

Specializing AI for Regulated Industries - How Domyn Uses NVIDIA Nemotron https://t.co/Zw8i7qU4gV

NVIDIA AI (@NVIDIAAI) — Tue, 08 Sep 2026 11:00:07 GMT

Specializing AI for Regulated Industries - How Domyn Uses NVIDIA Nemotron x.com/i/broadcasts/1…

💬14🔄8❤️88👀11660📊25

Congrats to our researchers for exceeding the gold medal threshold on the International Olympiad in ...

NVIDIA AI (@NVIDIAAI) — Sat, 05 Sep 2026 00:28:00 GMT

Congrats to our researchers for exceeding the gold medal threshold on the International Olympiad in Informatics (IOI) 2026 problem set 🥇

Our fine-tuned Nemotron model scored 535.4 out of 600, as graded by the IOI team — higher than the top-scoring human participant.

The team competed unofficially in Uzbekistan, where the International Technical Committee supervised the human contestants. The model had no internet access and faced the same time limits and submission constraints, using the same platform in parallel with the official competition.

Read more in the technical report:arxiv.org/abs/2609.02849B

💬49🔄85❤️604👀60638📊139

Read the technical blog for the five guidelines: https://t.co/HaXwdiTYYB

NVIDIA AI (@NVIDIAAI) — Fri, 04 Sep 2026 17:29:13 GMT

Read the technical blog for the five guidelines: nvda.ws/3SSYxA7

💬3🔄3❤️35👀11211📊15

Need faster LLM inference without sacrificing accuracy? Speculative decoding can help. Choosing the...

NVIDIA AI (@NVIDIAAI) — Fri, 04 Sep 2026 16:47:25 GMT

Need faster LLM inference without sacrificing accuracy? Speculative decoding can help.

Choosing the right draft length and drafting method depends on your model, workload and hardware. We break down five practical guidelines for balancing throughput and latency.

Your browser does not support the video tag.

💬32🔄51❤️436👀36113📊96

Read the technical blog for the five guidelines: https://t.co/v0z51AdDJP

NVIDIA AI (@NVIDIAAI) — Fri, 04 Sep 2026 16:47:25 GMT

Read the technical blog for the five guidelines: nvda.ws/4d1Ivut

💬1🔄1❤️9👀1149📊2

DGX Spark Live: Perplexity Portable Computer Goes Local https://t.co/jYwLOCtKly

NVIDIA AI (@NVIDIAAI) — Wed, 02 Sep 2026 19:30:06 GMT

DGX Spark Live: Perplexity Portable Computer Goes Local x.com/i/broadcasts/1…

💬10🔄20❤️188👀73535📊44

The Open Secure AI Alliance is beginning its next chapter with the Linux Foundation. We helped form...

NVIDIA AI (@NVIDIAAI) — Wed, 02 Sep 2026 18:28:16 GMT

The Open Secure AI Alliance is beginning its next chapter with the Linux Foundation.

We helped form the alliance to bring organizations together around open, practical ways to secure AI systems. Great to see that work continue with the Linux Foundation.

The Linux Foundation@linuxfoundation

The Open Secure AI Alliance is now part of the Linux Foundation. 

Originally founded by @nvidia, the Open Secure AI Alliance will develop open source tools, shared standards, and defensive practices that help organizations inspect, audit, and secure AI systems.

Learn more: bit.ly/3UqS67S

💬24🔄56❤️406👀35715📊64

AITX Austin Hackathon Winners Spotlight https://t.co/JHdIq4f7GV

NVIDIA AI (@NVIDIAAI) — Wed, 02 Sep 2026 18:00:08 GMT

AITX Austin Hackathon Winners Spotlight x.com/i/broadcasts/1…

💬5🔄5❤️58👀11832📊12

When tested against the original recorded attack, the optimized open pipeline improved mean backtest...

NVIDIA AI (@NVIDIAAI) — Tue, 01 Sep 2026 17:38:51 GMT

When tested against the original recorded attack, the optimized open pipeline improved mean backtest detection from 16.5% to 41.9% across independently seeded sessions.

It brought together domain context, specialized models, tools and validation.

💬1🔄2❤️30👀6028📊5

The detection rules were tested against eight new attacks not used to create them. Three rules met ...

NVIDIA AI (@NVIDIAAI) — Tue, 01 Sep 2026 17:38:51 GMT

The detection rules were tested against eight new attacks not used to create them.

Three rules met all quality gates and caught all eight.

The setup paired Nemotron 3 Ultra for orchestration with a fine-tuned Nemotron 3 Super for detection writing and repair.

Read the technical breakdown: nvda.ws/4iE6JOX

💬1🔄0❤️28👀4448📊5

Can a security defense catch an attack it hasn’t seen before? We teamed with @CrowdStrike to evalua...

NVIDIA AI (@NVIDIAAI) — Tue, 01 Sep 2026 17:38:50 GMT

Can a security defense catch an attack it hasn’t seen before?

We teamed with @CrowdStrike to evaluate an offensive-defensive system built on its SafeMind agentic system, where AI agents simulate controlled attacks, turn telemetry into detection rules, then test them against new attack paths.

Here’s how it works 🧵

💬36🔄90❤️608👀69905📊166

Ask the Experts: NVIDIA NeMo Switchyard | Nemotron Labs https://t.co/tqm3KDQfHa

NVIDIA AI (@NVIDIAAI) — Tue, 01 Sep 2026 16:59:52 GMT

Ask the Experts: NVIDIA NeMo Switchyard | Nemotron Labs x.com/i/broadcasts/1…

💬7🔄11❤️71👀11649📊17

NVIDIA Dynamo in 5 Minutes by our very own @Vishakha2394: https://t.co/FhS1FS6n81

NVIDIA AI (@NVIDIAAI) — Fri, 28 Aug 2026 21:03:32 GMT

NVIDIA Dynamo in 5 Minutes by our very own @Vishakha2394:
youtube.com/watch?v=mXYFcz…

💬1🔄2❤️18👀8957📊4

Already running an inference engine? So where does NVIDIA Dynamo fit in? In five minutes, we break ...

NVIDIA AI (@NVIDIAAI) — Fri, 28 Aug 2026 21:03:31 GMT

Already running an inference engine? So where does NVIDIA Dynamo fit in?

In five minutes, we break down how Dynamo sits around engines like @sgl_project, @vllm_project and TensorRT-LLM to scale inference across GPUs and nodes.

Full video in the comments 🔽

Your browser does not support the video tag.

💬53🔄40❤️325👀30450📊100

10 million downloads for NVIDIA Warp 🎉 Warp started with a simple idea: you shouldn’t have to leav...

NVIDIA AI (@NVIDIAAI) — Fri, 28 Aug 2026 15:31:57 GMT

10 million downloads for NVIDIA Warp 🎉

Warp started with a simple idea: you shouldn’t have to leave Python to get real GPU performance for physics and simulation.

Since then, developers have used it to accelerate work across physics simulation, computational engineering, geometry processing and robotics.

Thank you to everyone who downloaded it, broke it, filed an issue or sent a PR. On to the next 10M!

💬23🔄63❤️489👀27025📊92

📺 Tune into next week's livestream to learn how to accelerate computational engineering and robotic...

NVIDIA AI (@NVIDIAAI) — Fri, 28 Aug 2026 15:31:57 GMT

📺 Tune into next week's livestream to learn how to accelerate computational engineering and robotics with NVIDIA Warp:addevent.com/event/l123h7yl…3

💬0🔄1❤️10👀3066📊2

Run Your Polars Code on Multiple GPUs | Live with cuDF Polars https://t.co/QiZpoYAZ7l

NVIDIA AI (@NVIDIAAI) — Fri, 28 Aug 2026 14:02:00 GMT

Run Your Polars Code on Multiple GPUs | Live with cuDF Polars x.com/i/broadcasts/1…

💬12🔄5❤️75👀12707📊19

Congrats to @Alibaba_Qwen on releasing Qwen3.8-Flash-Next, an experimental open-weight model that pr...

NVIDIA AI (@NVIDIAAI) — Wed, 26 Aug 2026 17:17:05 GMT

Congrats to @Alibaba_Qwen on releasing Qwen3.8-Flash-Next, an experimental open-weight model that previews the Qwen4 architecture.

We’ve got Day 0 support to fine-tune with NVIDIA NeMo AutoModel and NeMo RL, plus recipes to run it with @sgl_project, @vllm_project and @lightseekorg TokenSpeed.

Get started: nvda.ws/4wV7yXp

💬29🔄53❤️612👀122146📊99

When an LLM engine crashes, a cold restart can mean minutes of lost capacity. Shadow engine recover...

NVIDIA AI (@NVIDIAAI) — Tue, 25 Aug 2026 21:59:12 GMT

When an LLM engine crashes, a cold restart can mean minutes of lost capacity.

Shadow engine recovery, a new preview feature in NVIDIA Dynamo, keeps a standby engine warmed up and ready to take over.

In our GLM-5.2 test, it restored capacity in 7.3 seconds, nearly 39x faster than a cold restart.

Read all about it: nvda.ws/45OhcA0

💬37🔄54❤️649👀62463📊124

Get Started with Open Model Routing | Nemotron Labs https://t.co/9Escn0IiWu

NVIDIA AI (@NVIDIAAI) — Tue, 25 Aug 2026 17:59:47 GMT

Get Started with Open Model Routing | Nemotron Labs x.com/i/broadcasts/1…

💬12🔄15❤️100👀11449📊24

Love seeing Nemotron 3.5 Lightning land in the top 4 open-weight models on @pinchbench with an 86.4%...

NVIDIA AI (@NVIDIAAI) — Mon, 24 Aug 2026 19:32:42 GMT

Love seeing Nemotron 3.5 Lightning land in the top 4 open-weight models on @pinchbench with an 86.4% avg success rate on standardized @OpenClaw agent tests 🦞

And Nemotron 3 Ultra is still holding down the #1 spot.

💬18🔄25❤️223👀15704📊42

https://t.co/3GN1Mz1j7M

NVIDIA AI (@NVIDIAAI) — Fri, 21 Aug 2026 17:21:39 GMT

x.com/NVIDIAAI/statu…

NVIDIA AI@NVIDIAAI

As AI agents take on more complex work, where should security live?

The harness guides what an agent tries to do. Infrastructure controls what it can actually do.

Our AI safety and security teams share how they’re thinking about security across the agent stack. 👇

💬1🔄5❤️102👀26576📊11

A closer look at how the stack fits together, from the components that guide an agent’s behavior to ...

NVIDIA AI (@NVIDIAAI) — Fri, 21 Aug 2026 16:43:55 GMT

A closer look at how the stack fits together, from the components that guide an agent’s behavior to the infrastructure that enforces its security boundaries.

Read the deep dive: nvda.ws/4zyvjHm

💬5🔄9❤️55👀6699📊15

As AI agents take on more complex work, where should security live? The harness guides what an agen...

NVIDIA AI (@NVIDIAAI) — Fri, 21 Aug 2026 16:43:54 GMT

As AI agents take on more complex work, where should security live?

The harness guides what an agent tries to do. Infrastructure controls what it can actually do.

Our AI safety and security teams share how they’re thinking about security across the agent stack. 👇

💬53🔄57❤️420👀74616📊107

NVIDIA AVO continuously inspects, plans, implements, and evaluates, using memory, tools, and executi...

NVIDIA AI (@NVIDIAAI) — Fri, 21 Aug 2026 13:05:09 GMT

NVIDIA AVO continuously inspects, plans, implements, and evaluates, using memory, tools, and execution feedback to build on what it learns along the way. This allows the system to sustain progress across long-running tasks rather than starting over with each model context.

Read about AVO and how we built it for long-horizon autonomous agents: nvda.ws/3S5qdS1

💬34🔄66❤️838👀148535📊187

Our general-purpose coding agent just scored 100% on the ARC-AGI-3 interactive reasoning benchmark. ...

NVIDIA AI (@NVIDIAAI) — Fri, 21 Aug 2026 13:01:03 GMT

Our general-purpose coding agent just scored 100% on the ARC-AGI-3 interactive reasoning benchmark.

NVIDIA AVO completed all 183 levels across all 25 public environments, figuring out what to do with no instructions, explicit rules, or stated goals.

💬339🔄702❤️6180👀1066576📊1386

https://t.co/lpXNR8BhIS

NVIDIA AI (@NVIDIAAI) — Thu, 20 Aug 2026 23:49:20 GMT

x.com/i/article/2090…

💬18🔄14❤️124👀15531📊34

Cosmos 3 Post-Training in Action With Aigen and Linker Vision | Cosmos Labs https://t.co/9pl5eAfX5M

NVIDIA AI (@NVIDIAAI) — Thu, 20 Aug 2026 16:01:34 GMT

Cosmos 3 Post-Training in Action With Aigen and Linker Vision | Cosmos Labs x.com/i/broadcasts/1…

💬10🔄10❤️55👀13004📊19

Check out NVIDIA cuOpt on GitHub and let us know what you build with it: https://t.co/T8DbCZeY91

NVIDIA AI (@NVIDIAAI) — Wed, 19 Aug 2026 21:52:51 GMT

Check out NVIDIA cuOpt on GitHub and let us know what you build with it: github.com/NVIDIA/cuopt

💬1🔄2❤️11👀4309📊3

NVIDIA cuOpt is the fastest open source solver on Hans Mittelmann benchmarks across three optimizati...

NVIDIA AI (@NVIDIAAI) — Wed, 19 Aug 2026 21:46:44 GMT

NVIDIA cuOpt is the fastest open source solver on Hans Mittelmann benchmarks across three optimization problem classes.

💬12🔄24❤️143👀14312📊30

Read the technical deep dive on SkillEvaluator: https://t.co/O1E9BZOOw7

NVIDIA AI (@NVIDIAAI) — Wed, 19 Aug 2026 16:28:18 GMT

Read the technical deep dive on SkillEvaluator: developer.nvidia.com/blog/evaluatin…

💬1🔄5❤️65👀9927📊20

We benchmarked 300+ NVIDIA verified skills to see how much they actually help agents on real tasks. ...

NVIDIA AI (@NVIDIAAI) — Wed, 19 Aug 2026 16:28:17 GMT

We benchmarked 300+ NVIDIA verified skills to see how much they actually help agents on real tasks.

Same task, same model, same setup. The only difference was whether the agent had the skill.

Across the benchmarks, skills improved correctness by 41 points, effectiveness by 39, and efficiency by 35.

SkillEvaluator is open source if you want to test your own skills before you ship them.

💬32🔄121❤️1048👀301870📊302

What are you building with open models? Show us what you’re working on and you could be headed to #N...

NVIDIA AI (@NVIDIAAI) — Tue, 18 Aug 2026 18:31:53 GMT

What are you building with open models? Show us what you’re working on and you could be headed to #NVIDIAGTC Berlin. 🎫   

Your Golden Ticket includes: 
→ Free conference pass 
→ VIP seating for Jensen’s keynote 
→ Exclusive NVIDIA merch 
→ Access to special events  

Submissions are open Aug 18 – Sep 10, 2026.  

Details and how to enter:nvda.ws/4wFdunbo

💬20🔄33❤️216👀23734📊60

Ask the Experts: What's New in the Nemotron Open Family | Nemotron Labs https://t.co/9aDHece5pA

NVIDIA AI (@NVIDIAAI) — Tue, 18 Aug 2026 18:00:07 GMT

Ask the Experts: What's New in the Nemotron Open Family | Nemotron Labs x.com/i/broadcasts/1…

💬9🔄15❤️67👀11177📊17

We just released TensorRT Model Connect in Public Preview. You can take a supported @huggingface mo...

NVIDIA AI (@NVIDIAAI) — Tue, 18 Aug 2026 16:24:46 GMT

We just released TensorRT Model Connect in Public Preview.

You can take a supported @huggingface model to end-to-end TensorRT inference in just two commands. No intermediate ONNX export, and the resulting bundle can run through native C++ APIs.

We also built the entire project with @OpenAIDevs Codex agents, with humans directing and reviewing the work. That includes model implementations, performance tuning, tests, integrations, and docs.

It’s open source, so go try it out, dig into the implementations, or contribute support for a new model: github.com/NVIDIA/TensorR…

💬33🔄67❤️539👀94295📊121

True. Favorite open source drop of the summer?

NVIDIA AI (@NVIDIAAI) — Tue, 18 Aug 2026 12:43:50 GMT

True. Favorite open source drop of the summer?

Ahmad@TheAhmadOsman

This summer has been a Homerun for Opensource AI and we’re only getting started btw

💬82🔄12❤️399👀42541📊82

Learn more: https://t.co/O0OtPGuYHj

NVIDIA AI (@NVIDIAAI) — Fri, 14 Aug 2026 18:59:43 GMT

Learn more: nvda.ws/457eK7l

💬0🔄2❤️9👀4021📊3

ICYMI: Not every step in an agent workflow needs the same model. Meet NVIDIA NeMo Switchyard, a n...

NVIDIA AI (@NVIDIAAI) — Fri, 14 Aug 2026 18:59:20 GMT

ICYMI: Not every step in an agent workflow needs the same model.  

Meet NVIDIA NeMo Switchyard, a new open-source library for model routing.

Use frontier models for complex reasoning and planning, and NVIDIA Nemotron Lightning for high-volume, specialized execution.

Your browser does not support the video tag.

💬16🔄25❤️166👀16064📊39

@AIatMeta Demo + cookbook if you want to try it out yourself 👇 Platform cookbook: https://t.co/XeD...

NVIDIA AI (@NVIDIAAI) — Thu, 13 Aug 2026 18:01:49 GMT

@AIatMeta Demo + cookbook if you want to try it out yourself 👇

Platform cookbook:github.com/meta-models/me…Y
Demo/recipe:github.com/meta-models/me…w

💬1🔄2❤️17👀4360📊4