Teaching your AI new tricks.
Santa Clara, CA
- Tweets 13,988
- Following 900
- Followers 353,173
- Likes 10,302
Europe’s industrial future is taking shape at GTC Berlin.
With our CUDA-X and Omniverse libraries, plus Cosmos and Nemotron open models and tools, Europe’s industrial leaders are testing designs faster, simulating robots and improving factory operations.
Learn how Agile Robots, @Audi, @Dassault3DS, Schaeffler and @Siemens are putting these technologies to work: nvda.ws/4BnYgGT
17
23
165
10,348
AI agents need to know your business to be useful.
At GTC Berlin, we’re adding NVIDIA Agent Knowledge Foundation to NVIDIA Agent Toolkit. It helps agents find relevant information, make predictions and help people decide what to do next.
OpenShell gives teams control over what agents can access and do.
Learn more: nvda.ws/4zJtVRh
19
12
100
9,405
We’re releasing more open models to help teams across Europe build AI with their own data, on infrastructure they control.
Announced at GTC Berlin, the models help developers, researchers and enterprises build specialized AI agents, train robots, design proteins and develop quantum computers.
See what’s new: nvda.ws/4iwBMvF
30
42
325
21,773
AI agents need fast tool calls and security controls they can’t rewrite.
NVIDIA OpenShell Reference Architecture combines the compute to run agents at scale with runtime controls that govern their access to tools and data.
See how the hardware and software work together: nvda.ws/4BiJBfY
34
31
243
15,061
Doors just opened - countdown is on! Watch the live pre-note piped.video/live/PrA_ayyy22E…
15
22
205
30,635
We’re working with the European Broadcasting Union and @Adobe to help media teams keep control of their data and find the footage they need faster.
With the EBU, we’re building on sovereign AI standards so broadcasters can use AI while keeping control of their data and protecting editorial independence.
Adobe is using our GPUs to make Frame.io video search faster, helping TV and film producers find a specific scene in years of footage.
Read about both announcements from GTC Berlin: nvda.ws/4sheBsJ
9
21
132
10,884
Berlin is an epic location for @NVIDIAGTC!
Follow along for announcements from Jensen's keynote tomorrow.
24
36
522
20,522
That's a wrap on Build-A-Claw Berlin!
And we of course served @lobsterlager. A father/son duo used @openclaw to make a lager.
Thank you to everyone who stopped by and we're looking forward to an amazing week at @NVIDIAGTC Berlin.
21
34
234
20,132
NemoClaw was just named one of TIME’s Best Inventions of 2026! 🦞
So proud of the team and shoutout to everyone building safer AI agents with us.
Blurb on @TIME: time.com/collection/best-inv…
These are creations changing how we live, work, play, and think about what’s possible. time.com/article/2026/10/15/…
18
23
178
13,952
And this one fits into 24GB of memory.
Introducing Nemotron 3.5 Bolt 24GB ⚡️
NVIDIA Nemotron 3.5 Bolt 24GB puts powerful local agents within reach. ⚡
74B-A9B parameters packed into 24GB of memory, balancing intelligence and speed while running alongside your apps. Optimized for RTX Spark and ready for your next project.
66
104
1,633
100,276
Nemotron 3.5 Super VL pairs strong reasoning with high output speed, even at 100K input tokens. See how it compares on @ArtificialAnlys 👇
3
8
112
15,346
Congrats to StepFun on Step 5 Preview 🙌 Their largest open weights release.
The 600B MoE model is built to handle agent tasks lasting up to 24 hours.
The model runs at nearly 8K tokens/sec/GPU on NVIDIA GB300 using NVFP4.
NVP4 checkpoint: huggingface.co/stepfun-ai/St…
Deploy with @SGL_project - docs.sglang.io/cookbook/auto… or @vLLM_project - recipes.vllm.ai/stepfun-ai/S…
Post-train with NeMo Automodel: github.com/NVIDIA-NeMo/Autom…
Run it with NVIDIA NIM using a single command: catalog.ngc.nvidia.com/orgs/…
Step 5 Preview is now open weights.
BF16 and FP8 weights are live on Hugging Face and ModelScope, together with our technical blog.
• Sparse MoE: 600B total, 27B active per token
• 1M-token context window
• Native text and vision input
• Built for long-horizon agentic work: coding, software engineering, professional knowledge work and finance
Frontier intelligence, open weights. Build what's next.
Hugging Face: huggingface.co/stepfun-ai/St…
ModelScope: modelscope.ai/models/stepfun…
GitHub: github.com/stepfun-ai/Step-5…
34
33
412
19,088
Build-A-Claw London ✅️
A visit from @JensenHuang ✅️
A very good day to have a raffle ticket ✅️
Thank you to everyone who stopped by!
See you in the next city.
17
35
346
15,830
We just shared how we post-trained Nemotron for this result, plus IMO gold-level performance.
Check out the recipe 👉 huggingface.co/blog/nvidia/n…
3
19
2,680
Barcelona showed up ready to build 🦞.
The road to #NVIDIAGTC Berlin is officially underway.
6
21
165
21,781
An AI agent makes a mistake early in a task, then keeps going in the wrong direction.
Our researchers built PivotOPD to teach agents how to avoid those mistakes and recover when they happen. During training, a teacher model shows the agent a better action and how to get back on track over the next few steps.
Read the paper and watch how it works: research.nvidia.com/labs/lpr…
67
62
492
39,344
Some telecom network problems don’t have a fix in the runbook.
AI agents can investigate, compare fixes and test them in simulation. NVIDIA Nemotron helps agents reason through problems, while NVIDIA OpenShell controls what they can access and do.
See how: nvda.ws/4y9vB5x
2
3
22
7,395
In our 2026 telecom AI survey:
• 89% of respondents said open source models and software are important to their AI strategy.
• 66% said their company is actively using AI, up from 49% last year.
Read the report: nvda.ws/4rPpi5i
2
1
17
4,589
Telcos need AI agents that understand their networks.
We provide the full stack to turn open models into specialized agents while keeping operators in control.
@ATT, @SoftBank and @Indosat are adopting open models for use cases like autonomous network operations and adapting them to local languages and industry needs.
Here’s why they’re choosing open models and how they’re putting them to work. 🧵
Read more: nvda.ws/4dHMAEL
17
9
64
12,495
Huge thank you to everyone who downloaded Nemotron 3 Diarization and helped it trend on @huggingface! And we appreciate all the comments.
@sabbassi_11 answered a few of your questions:
When several people talk at once, a transcript can get messy fast.
Our new Nemotron 3 Diarization model tracks who spoke when, even when voices overlap. It handles up to eight speakers, has 100M parameters, and is now available on @huggingface 🤗
20
15
122
22,025
A speech model can support Arabic and still struggle with local dialects.
Fine-tuning NVIDIA Nemotron 3.5 ASR cut word error rate on Najdi and Hijazi Saudi Arabic from 55% to 30%.
Check out our new tutorial and learn how to adapt it for other dialects and languages: nvda.ws/4hYqGPM
137
297
2,763
680,206
We’re looking for our next cohort of graduate fellows.
Our 2027–2028 Graduate Fellowship Program is open to Ph.D. students worldwide with awards up to $60,000, plus mentorship and technical support.
Apply by October 30: nvda.ws/4rGKDhv
10
108
596
116,625
We’re going live tomorrow at 9 a.m. PT to build a visual AI agent from a single prompt. Bring your questions piped.video/watch?v=PQJKs1dy…
1
1
14
6,353
One prompt built and deployed a visual AI agent for a manufacturing line in under 30 minutes, with alerts, video search, and incident reports.
Check out our new tutorial that shows you how to build one with the new Build Vision AI skill in the NVIDIA VSS Blueprint 3.3.
Tutorial: nvda.ws/4AEyKfS
33
53
424
28,673
Your agent got the right answer. How much work did it take to get there?
We worked with @NousResearch/@Teknium on a hands-on walkthrough of NVIDIA NeMo Relay to collect traces for Hermes Agent. Run two example scenarios, follow the agent’s calls and retries, and see the full trace in Arize Phoenix.
The blog also looks at how Nous used traces and task results to evaluate fixes across repeated runs.
TechBlog: nvda.ws/47tKxjQ
📽️ by our own @Picomoorhead:
36
47
419
86,684
Physis-Lang with NVIDIA Cosmos 3 Super and Nano rank #1 and #2 on the Physics-IQ Verified image-to-video leaderboard.
The benchmark tests how well generated videos follow real-world physics.
Full rankings: physics-iq-verified.anates.a…
5
22
4,441
A video world model can make a convincing clip and still get the physics wrong.
Our researchers just released Physis-Lang, an open self-evolving framework that adds physics reasoning to video captions. The captions explain why and how a scene unfolds. We use them to fine-tune world models and add that reasoning to prompts when generating video.
Adding physics reasoning to the prompt alone improved NVIDIA Cosmos 3’s PhyGenBench score by 5.62 points without retraining.
Paper and project: physis-intelligence.github.i…
22
81
457
31,763
AI agents need clear limits on what they can do, and those limits need to hold up while they’re working. @JensenHuang was on CNBC this morning to talk about the safety measures we’re building to help make that happen.
🎥 from @SquawkCNBC:
84
92
528
54,936
Nemotron 3 Diarization ranked #1 of 12 systems in @voicearena_ai's initial Diarization-Bench results. Its 14.72% error rate was ~24% lower than the runner-up.
Here’s a four-speaker comparison from the benchmark.
Full results: voicearena.com/diarization-b…
6
13
288
55,668
When several people talk at once, a transcript can get messy fast.
Our new Nemotron 3 Diarization model tracks who spoke when, even when voices overlap. It handles up to eight speakers, has 100M parameters, and is now available on @huggingface 🤗
207
862
9,710
1,067,009
Figuring out a data center’s carbon footprint means piecing together a lot of messy data, from parts lists and documents to supplier emails.
Sluicebox built an AI supplier agent called Lucy to help. Using NVIDIA Nemotron 3 Ultra, Lucy collects and checks supplier data so engineers can understand the impact of their choices while they’re still designing.
In Sluicebox’s tests, Nemotron improved accuracy, cut costs by 51–80% and delivered responses up to 2x faster.
See how it works: nvda.ws/4iIQi4e
19
34
195
18,391
Your LLM endpoint works. But how does it perform when traffic increases?
NVIDIA Dynamo AIPerf helps you measure TTFT, ITL, latency and throughput at scale, then test with realistic traffic patterns you can reliably repeat.
Read the blog: nvda.ws/4hdIeGd
50
60
461
50,778
Images rarely show an object’s full 3D geometry.
At #ECCV2026, our research team introduced Axolotl3D, a multimodal and occlusion-aware 3D generation model. It combines images, camera data and partial geometry to reconstruct missing regions while preserving observed ones, achieving state-of-the-art results across single- and multi-view settings.
Project page: research.nvidia.com/labs/sil…
28
55
478
31,460
How can a 30B-parameter model activate just 3B parameters per token and still draw on the full model’s capacity?
Learn how dense and MoE models use parameters differently, and what that means for throughput, memory and serving complexity.
Check out our new technical explainer: nvda.ws/4Aaix21
23
30
199
19,208
News teams can help verify whether footage might be AI generated. Sports producers can create smoother slow-motion replays. Broadcasters can translate programming with lip-synced dubbing.
At #IBC2026, we added new SDKs, NIM microservices, playbooks and blueprints to the NVIDIA AI for Media collection.
See what's new: nvda.ws/469nQB0
15
15
116
13,853
A lot of work goes into serving a model efficiently.
For the Nemotron 3 Ultra NIM, our engineers tuned caching, memory, parallelism, decoding and more. On four B200 GPUs, those optimizations supported up to 2.5x more concurrent users while maintaining 50 TPS/user.
Read the engineering deep dive → nvda.ws/4xX7SWU
ALT Graph showing Nemotron 3 Ultra NIM serving up to 2.5x more users.
32
37
350
24,978