This account is inactive. Follow @NVIDIAAI for the latest updates.
Santa Clara, CA
- Tweets 7,209
- Following 387
- Followers 111,536
- Likes 9,384
From interactive hair simulation to an #AI system that can auto-generate realistic 3D character motion by watching 2D videos, check out the 19 papers from #NVIDIAResearch accepted into #SIGGRAPH2023.
Read the blog with links to each paper: nvda.ws/3nniRKc
61
672
3,331
458,644
Introducing DeepSeek-R1 optimizations for Blackwell, delivering 25x more revenue at 20x lower cost per token, compared with NVIDIA H100 just four weeks ago.
Fueled by TensorRT DeepSeek optimizations for our Blackwell architecture, including FP4 performance with state-of-the-art production accuracy, it scored 99.8% of FP8 on MMLU general intelligence benchmark.
FP4-optimized DeepSeek checkpoint now available on @huggingface: huggingface.co/nvidia/DeepSe…
105
415
2,851
501,049
✨Introducing Neuralangelo✨ A new #AI model to turn 2D video from any device -- cell phone to drone capture -- into 3D structures with intricate details.
New from #NVIDIAResearch #CVPR2023
Read more: nvda.ws/3OPOlDR
59
479
2,048
497,928
Let us introduce you to the next evolution of #GauGAN:
#GauGAN360 turns your simple sketch into a 360 degree HDR environment. Try it now on our interactive #AI Playground. nvidia.com/en-us/research/ai…
22
358
1,485
Out of the box, and now powering up in the lab.
Stay tuned …. #SparkSomethingBig
68
113
1,287
148,449
Securely experiment and build your own specialized agents, as the 671-billion-parameter DeepSeek-R1 model is now available as an NVIDIA NIM microservice in preview on build.nvidia.com.
Learn more ➡️ nvda.ws/4grQaBq
110
248
1,119
643,356
🎉 NVIDIA has achieved the world’s fastest DeepSeek-R1 inference.
👀 A single system with eight Blackwell GPUs in an NVL8 configuration delivers 253 TPS/user or 30K TPS system throughput on the full DeepSeek-R1 671B parameter model.
#GTC25 ➡️ nvda.ws/3FzAzCO
35
152
998
244,206
Trained in minutes, rendered in seconds. ⏱️
Our Instant NeRF from NVIDIA Research turns 2D photos into 3D scenes in the blink of an #AI. #nerfies #neuralnetworks #SX70 #GTC22 nvda.ws/36HL5GB
16
253
985
Have you tried the new quick deploy feature in NGC? In this new video, @drahmadbazzi takes it for a spin and shows how to deploy an object detection #jupyternotebook on @googlecloud's Vertex #AI Workbench. nvda.ws/3xygGoT
85
50
542
Yes, it really does work! #notphotoshopped. Watch the #GTC20 keynote to learn how its done: piped.video/04E3EjsQLYo?t=721
If this works, it’s just nuts. Nvidia has come up with a video compression algorithm that lets you do Zoom calls with a tiny fraction of the bandwidth usually required. Could be a game-changer for people with poor internet connections. piped.video/watch?v=NqmMnjJ6… (Via @chrismessina!)
22
208
877
To generate a seemingly infinite number of portraits in a variety of painting styles, @NVIDIA researchers developed StyleGAN2. The new model, trained on NVIDIA V100 GPUs with @TensorFlow, will be presented at #CVPR2020. Learn more here: nvda.ws/2UJ3udu
13
233
861
Unveiled at #NeurIPS2018, NVIDIA trained a conditional generative neural network on existing video to create the first ever interactive neural network render. #AI rendering virtual worlds in real-time. Learn more:
nvda.ws/2rfmPDW
14
266
816
Is your game avatar just not understanding you?
With the new #AI framework, PADL, you can more easily command your 3D characters with natural language commands. Learn more about this state-of-the-art paper here: nvda.ws/3DuIwFB
8
174
855
137,031
A preview Into the future of #quantum-GPU computing:
At GTC Washington, D.C., we announced NVIDIA NVQLink, an open system architecture for tightly coupling the extreme performance of GPU computing with quantum processors to build accelerated quantum supercomputers.
In the expo hall, accelerated quantum computing was a centerpiece of the NVIDIA booth.
See our live blog: blogs.nvidia.com/blog/nvidia…
33
169
847
55,257
We're calling it Super SloMo. See how @NVIDIA researchers developed a #deeplearning method that uses #GPUs to turn any video into a slow motion masterpiece. #CVPR2018 nvda.ws/2Mr2ZPt
21
258
804
Check out this #AI mobile app called @MugLifeApp trained on #GPUs that lets you bring photos to life with 3D animation. nvda.ws/2zo2ov3
21
452
800
We're calling it Image Inpainting. See how @NVIDIA researchers introduced a state-of-the-art #deeplearning method that uses #GPUs to edit images or reconstruct a corrupted image in real-time. nvda.ws/2K1G243
25
284
773
Announcing NVIDIA #TITANRTX: the world’s most powerful desktop GPU. Adds massive performance for AI research, data science and creative applications - on display this week at #NeurIPS2018. Learn more: nvda.ws/2rj6HRL
43
186
749
We just released 3 million samples of high quality vision language model training dataset for use cases such as:
📄 optical character recognition (OCR)
📊 visual question answering (VQA)
📝 captioning
🤗 Learn more: nvda.ws/4oyfevu
📥 Download: nvda.ws/4fz2gtB
28
155
806
95,745
Take a ride with NVIDIA research in the first ever interactive neural network renderer displayed at #NeurIPS2018. Read to learn more about this incredible work: nvda.ws/2G6whE6
5
81
626
What happens when you combine the capabilities of @openAI's ChatGPT and #NVIDIAOmniverse?
You unlock the power of #GenerativeAI. 🔓
➡️ bit.ly/3nOS8Gh
12
159
630
129,828
Ready to build an AI agent?
Our new guide covers everything from agent architecture to state management with LangGraph, all powered by NVIDIA Nemotron on @openrouter.
Learn advanced concepts and get started instantly with a one-click NVIDIA Launchable.
🔗 nvda.ws/468NmHt
12
75
638
174,507
🏎️ ️We accelerated @OpenAI’s new GPT open weight models -- gpt-oss-20b and gpt-oss-120b -- for leading inference performance on NVIDIA Blackwell architecture, delivering up to 1.5 million tokens per second on an NVIDIA GB200 NVL72 system.⏱️🏁
👀 Technical deep dive on how we maximized model performance with the OpenAI community including @huggingface, @ollama, and @vllm_project so you can deploy them from the Cloud to the Edge.
Tech Blog ➡️ nvda.ws/4m0KzFs
23
90
628
46,265
Deep learning-based 3D shape reconstruction is an open problem in AI. “Deformable Tetrahedral Meshes” presented @ #NeurIPS2020 by NVIDIA, is the first approach to showcase high-quality 3D tetrahedral meshes using only a single image as input. Learn more: nvda.ws/3gAvdXr
3
131
573
📣 Announcing the release of OpenReasoning-Nemotron: a suite of reasoning-capable LLMs which have been distilled from the DeepSeek R1 0528 671B model. Trained on a massive, high-quality dataset distilled from the new DeepSeek R1 0528, our new 7B, 14B, and 32B models achieve SOTA perf on a wide range of reasoning benchmarks for their respective sizes in the domain of mathematics, science and code. The models are available on @huggingface🤗: nvda.ws/456WifL
12
115
602
55,701
👀 Accelerate performance of @AIatMeta Llama 4 Maverick and Llama 4 Scout using our optimizations in #opensource TensorRT-LLM.⚡
✅ NVIDIA Blackwell B200 delivers over 42,000 tokens per second on Llama 4 Scout, over 32,000 tokens per seconds on Llama 4 Maverick.
✅ 3.4X more performance and 2.6X lower cost compared to Hopper H200.
61
87
585
75,735
🏗️ Build #3D scenes from a single input view based on the latest #NVIDIAResearch paper accepted at #ICCV2023.
👀 This 3D-aware #generativeAI model can build diverse and plausible views.
research paper: nvda.ws/3OkfYmX
project page: nvda.ws/3Krkxuy
14
145
557
113,879
👀 Reserve your NVIDIA DGX Spark -- A Grace Blackwell AI supercomputer on your desk, capable of delivering 1000 TOPS with 128GB unified memory for inferencing up to 200B parameters.
Prototype, fine-tune, and deploy with ease. 🙌
➡️ nvda.ws/42PjjTy
#SparkSomethingBig 💫
41
90
555
59,902
Amazing! Researchers from @UCSF used #AI + @NVIDIA GPUs to give a paralyzed man the ability to communicate by translating his brain signals into computer generated writing. Read more: nvda.ws/3exANKm
16
123
530
8x longer context using 50% less VRAM with FlexAttention!
👏 Congrats @UnslothAI on this amazing performance with OpenAI's gpt-oss and NVIDIA GPUs
OpenAI gpt-oss with ultra long context is here!🚀
Introducing Unsloth Flex Attention which enables 61K context for gpt-oss bf16 training on a 80GB GPU.
Unsloth achieves 8×longer context, 50% less VRAM & 1.5×faster training vs. all implementations.
🔗docs.unsloth.ai/basics/long-…
4
81
556
36,560
Our Llama-3.1-Nemotron-70B-Instruct model is a leading model on the 🏆 Arena Hard benchmark (85) from @lmarena_ai.
Arena Hard uses a data pipeline to build high-quality benchmarks from live data in Chatbot Arena, and is known for its predictive ability of Chatbot Arena Elo score as well as separability between helpful and less helpful models.
Use our customized model Llama-3.1-Nemotron-70B to improve the helpfulness of LLM generated responses in your applications.
📥 Try on our API catalog: build.nvidia.com/nvidia/llam…
📥 On @GitHub: github.com/lmarena/arena-har…
📥 Or on @huggingface: huggingface.co/nvidia/Llama-…
30
122
502
140,756
It's being called a game changer in animal tracking! This #AI was trained on @NVIDIA #GPUs to track the body parts of multiple species automatically. nvda.ws/2P3CqjH
14
105
485
NVIDIA GET3D is a generative #AI model that can be used to populate virtual worlds with 3D objects and characters. #GTC22
Learn more: nvda.ws/3UzNS9B
8
110
499
Wondering what color you should go with? @ModiFace's new #AI app trained on GPUs colorizes hair in real-time. nvda.ws/2xwg3Mo
18
214
486
🎊 Our CEO Jensen Huang unveiled our new NVIDIA Project DIGITS personal AI supercomputer which enables developers across industries to prototype, fine-tune, and inference AI models on their desktop. 👀
📹 piped.video/live/k82RwXqZHY8…
➡️ nvidia.com/en-us/project-dig…
Deploy locally and seamlessly scale to the data center or cloud. Powered by the NVIDIA GB10 Superchip and based on #NVIDIAGraceBlackwell. #CES2025
56
80
473
91,915
Today, NVIDIA and @huggingface announced that #NVIDIADGX Cloud will integrate into the Hugging Face platform, empowering millions of developers with #generativeAI supercomputing for large language models and advanced #AI applications. #SIGGRAPH2023
➡️ nvda.ws/45bK1UM
14
101
463
96,726
“Researchers need open source. Developers need open source. Companies around the world — we need open source.” -Jensen Huang
#NVIDIAGTC
23
94
491
118,676
Get ready for an unforgettable keynote from NVIDIA CEO Jensen Huang, where he’ll unveil what’s next in agentic AI, robotics, and accelerated computing.
📍 October 28 at 12PM ET / 9AM PT, streaming live
🎬☕+ kick-start keynote day with the #NVIDIAGTC Live pregame show starting at 8:30AM ET.
Add to your calendar: nvda.ws/4oIjWX8
15
65
474
41,736
Create a personalized chatbot with the Chat with RTX tech demo. #AIonRTX
Quickly get tailored info from your own files and content by connecting your data to an #LLM on RTX-Powered PCs.
📥 GeForce RTX 30 and 40 series owners can download now: nvda.ws/48iOtlz
16
110
460
46,785
Announcing NVIDIA T4 Tensor Core GPU, the world’s most advanced #AI inference accelerator designed to deliver state-of-the-art #inference in real-time. #GTC18 nvda.ws/2p0fkQd
8
133
440
🎵 ✨The world’s most flexible sound machine?
With text and audio inputs, this new #generativeAI model, named Fugatto, can create any combination of music, voices, and sounds.🎹
Read more in our blog by @RichardKerris ➡️ blogs.nvidia.com/blog/fugatt…
#NVIDIAResearch
Note: Some music, sounds, and the voice of NVIDIA CEO Jensen Huang used in this video are AI generated.
49
163
461
90,346
While at #OpenSourceAIWeek, @Karpathy received one of our first NVIDIA DGX Sparks, personally delivered by @NaderLikeLadder and @Baxate_carter.
🙌 Excellent work by Andrej on introducing Nanochat -- an #opensource project that makes it easy for developers to explore and train their own LLMs from scratch.
We can’t wait to see what you’ll do with your Spark! #SparkSomethingBig ✨
6
28
460
48,849
🚨 GeForce RTX 5090s are available at the Gear Store in the park right now at #GTC25
90 units are available for the next 30 minutes, with more coming tomorrow. Come say hi!
36
81
452
67,918
👀 We just opened over 26M lines of synthetic data that was used to train the Llama Nemotron Super v1.5 model.
🔎 This transparency into our model training also helps you build your own models -- without expending the effort and time required to produce your own datasets.
🔢 Find them on @HuggingFace 🤗 huggingface.co/datasets/nvid…
9
76
449
38,792
🌟Developers - you heard it from Jensen - we want to hear your most creative ideas to rename #ProjectDIGITS, our latest AI supercomputer.
🙌 Share your ideas from profound to playful. #RenameProjectDIGITS 🍿
😃 How excited are you about putting this AI powerhouse into action?✨
133
77
417
88,986
No need for expensive motion-capture hardware.
#AI researchers from @NVIDIA developed a new #physics-based method to estimate and synthesize human motion models from 2D videos. Learn more: nvda.ws/324ov9p
2
97
431
⚡ 3.2x higher throughput in just two weeks for even more cost-effective DeepSeek-R1 #inference.
👀 Check out DeepSeek-R1 performance gains on NVIDIA Hopper GPUs, which feature powerful FP8 Tensor Cores and high-bandwidth NVLink and NVSwitch interconnects, running SGLang by @lmsysorg.
17
58
438
63,903
We’re excited to share that 🥇Llama Nemotron Super 49B v1.5 -- our latest open reasoning model -- is now #1 on the @ArtificialAnlys Intelligence Index - a leaderboard that spans advanced math, science, and agentic tasks, in the 70B open model category.
Llama Nemotron Super 49B v1.5 is trained with high-quality reasoning synthetic data generated from models like Qwen3-235B and DeepSeek R1. It delivers state-of-the-art accuracy and throughput, running on a single H100. Key features:
🎯 Leading accuracy on multi-step reasoning, math, coding, and function-calling
🏗️ Post-trained using RPO, DPO, and RLVR across 26M+ synthetic examples
📊 Fully transparent training data and techniques
If you're building AI agents and want a high accuracy, fully-open, and transparent reasoning model that you can deploy anywhere, try Super v1.5 on nvda.ws/4lLMYE1 or download from Hugging Face 🤗 nvda.ws/3U110Wf
➡️ Leaderboard: nvda.ws/41cs3Se
18
87
426
100,685
Is your 3D character phasing, glitching, or clipping through objects?
This #NVIDIAResearch #AI approach takes physics-based character motion generation and teaches the character how to realistically interact with objects & the environment. #SIGGRAPH2023
nvda.ws/44XodMl
10
93
392
59,671
⏱️ DYK you can pre-train your own model in just 4 hours with Nanochat?
We loved Nanochat from @karpathy so much that our engineers built a ready-to-go GPU-provisioned Launchable to easily train your own LLM. 👀 The first 10 devs to deploy will receive free access to NVIDIA H100 or L40S GPUs to train their own GPT-2–style model.
➡️ blogs.nvidia.com/blog/open-s…
#OpenSourceAIWeek
9
47
412
109,722
🎉Just released: Eureka!, a new AI agent that uses LLMs to automatically generate algorithms to train robots to accomplish complex tasks.
👀 The #NVIDIAResearch paper includes the AI algorithms and how to experiment with Eureka using NVIDIA Isaac Gym.
👇 nvda.ws/3PWKlk4
15
110
384
71,518
Director of Research, @AnimaAnandkumar, takes us behind the scenes of NVIDIA's new core #AI & machine learning research lab to share some challenges in algorithmic research her team hopes to tackle. Stop by to meet her in person next week at #NeurIPS2018. nvda.ws/2KJFcd1
1
86
383
At #NeurIPS2018, @datametrician shares how the new NVIDIA #TITANRTX provides a great entry point for developers and data scientists and the flexibility to experiment on larger datasets. Learn more about how to leverage TITAN RTX with @rapidsai: nvda.ws/2FVOWC7
5
65
349
The new NVIDIA Ampere architecture delivers the unmatched acceleration at every scale with faster performance for HPC, AI and data analytics workloads. Get an in-depth look on the NVIDIA Ampere Architecture: nvda.ws/2Wv2RWo #GTC20
7
91
343
With NVIDIA AI playground, you can now experience NeVA, a multi-modal vision language model, directly from your browser.
Just upload an image to NeVA model and ask questions related to the image. #generativeai nvda.ws/459gTxw
8
93
374
75,960
Watch how @OpenAI uses #deeplearning and @NVIDIA #GPUs to train robots to grasp and manipulate real-world objects with human-like dexterity. nvda.ws/2v3QBxV
3
135
355
🎉 Congratulations @IBM on the launch of your new Granite 4.0 family of open LLMs.
🙌 Granite 4.0’s design for maximum efficiency, trust and security is a perfect match for secure, production-ready deployment on any NVIDIA accelerated infrastructure with NIM.
Get started with Granite 4.0 -- now available on NVIDIA NIM ➡️ build.nvidia.com/nvidia/brin…
The next generation of open-source AI is here. Meet IBM Granite 4.0.
We've engineered these open, performant and trusted small language models to solve key enterprise challenges. Discover how Granite 4.0 delivers exceptional performance while requiring only a fraction of the resources: ibm.co/6019BJqJP
10
60
376
29,150
👀 The latest SGLang inference optimizations on DeepSeek-R1 represent a 26x increase in H100 performance in only 4 months. 📈
🎉 Congrats to the @lmsysorg team.
6
65
375
36,293
.@NVIDIA researchers just developed a #GPU-accelerated reinforcement learning simulator that can teach a virtual robot human-like tasks in record time. #Corl18 nvda.ws/2AB7a7t
8
129
347
📣 Announcing Llama Nemotron Super v1.5 📣
This release pushes the boundaries of reasoning model capabilities at the weight class of the model and is ready to power agentic applications from individual developers, all the way to enterprise applications.
📈 The Llama Nemotron Super v1.5 achieves leading reasoning accuracies for science, math, code, and agentic tasks while delivering up to 3x higher throughput.
Try it on build.nvidia.com, or download from @huggingface: 🤗 huggingface.co/nvidia/Llama-…
Tech blog: developer.nvidia.com/blog/bu…
10
72
359
32,380
Congrats @AnimaAnandkumar for receiving the 'VB Women in AI' award in the AI Research category. Fellow NVIDIA researchers @FidlerSanja, and @SifeiL, were also among the 21 leaders nominated for #VBTransform awards. nvda.ws/2Ce3Frj
16
42
350
Creativity unleashed 👨🎨🎨✨
👀 Edify3D from #NVIDIAResearch explores a new way to generate text-to-#3D images.
📗 Project page: research.nvidia.com/labs/dir…
📝 Paper: arxiv.org/abs/2411.07135
🔊 sound on
48
77
343
177,501
We are so honored. NVIDIA DGX Spark was just named one of @TIME's Best Inventions of 2025 🏆
Our new desktop AI supercomputer will deliver an unprecedented 1 petaflop of AI compute, right to your workspace, making state-of-the-art AI more accessible than ever. 🙌
We can’t wait to see what innovators build ... stay tuned for updates.
time.com/collections/best-in…
#SparkSomethingBig ✨
27
43
352
23,803
New York City-based startup @TheTake_AI, a member of @NVIDIA’s Inception program, recently unveiled a new #deeplearning algorithm that can automatically decode what a celebrity, athlete, or other public figure is wearing in a video in near real time. nvda.ws/2InkwJv
9
128
342
🎓 Ready to earn your NVIDIA Generative AI Certification?
Watch the on‑demand webinar to explore exam prep strategies, sample questions, and expert insights to help you ace the new professional‑level Agentic AI and Generative AI LLM certifications.
➡️ Watch now: nvda.ws/48Y27yz
9
43
344
20,597
🎊 Llama Nemotron Ultra 253B is here 🎊
✅ 4x higher inference throughput over DeepSeek R1 671B
🏆Highest accuracy on reasoning benchmarks:
💎 GPQA-Diamond for advanced scientific reasoning
💎 AIME 2024/25 for complex math
💎 LiveCodeBench for code generation and completion
Try as #NVIDIANIM ➡️ build.nvidia.com/nvidia/llam…
Technical deep dive ➡️ developer.nvidia.com/blog/bu…
12
63
341
85,022
A new @NVIDIA Jetson-based autonomous robot by @doxel_ai captures 3D scans of construction sites to inspect the quality of the work and tracks the progress of the project. nvda.ws/2rNVE6v
7
143
321
Trained on 50,000 episodes of PAC-MAN with @NVIDIA V100 GPUs, this new GAN model developed by NVIDIA Researchers from our Toronto lab, generates a fully functional version of the dot-munching classic without an underlying game engine, just #AI #pacman40th nvda.ws/2LPBsZ6
8
79
333
⚡ We just accelerated output of @OpenAI’s gpt-oss-120B by nearly 2x 👀
In collaboration with @ArtificialAnlys, we demonstrated impressive performance of OpenAI's gpt-oss-120B on a DGX B200 system with 8xB200:
⚡️ Nearly 900 output tokens/s in single query tests
⚡️ Close to 600 output tokens per second per user for 10 users at once — about 6,000 total tokens/second
Using TensorRT-LLM with speculative decoding, our #NVIDIABlackwell hardware natively supports FP4 precision. Single B200 DGX deployments achieved 900 tokens/s in single query tests.
We are currently working with many of our end-point partners to run more tests on NVIDIA H100, H200, and B200 systems with Artificial Analysis. Stay tuned for results…
17
49
346
29,120
NeMo Guardrails, the new toolkit for easily developing trustworthy LLM-based conversational applications.
It works with all #LLMs and natively supports @langchain, adding a layer of safety, security, and topical guardrails to existing applications. ➡️ bit.ly/3LpW3mi
ALT Available Now: NVIDIA NeMo Guardrails for LLMs
5
79
327
161,671
We just launched new open models and datasets to make AI research and development more accessible 🤝
You now have open foundations to build specialized intelligent agents faster, safer, and at scale — from Nemotron to Cosmos, Isaac GR00T to Clara.
Over 650 open models and 250 open datasets contributed by NVIDIA are now available on @huggingface.
Learn more → nvda.ws/4hwUVvB
11
69
335
25,255
A new #GenerativeAI method by NVIDIA researchers uses off-the-shelf, pre-trained latent diffusion models (LDMs) to turn image generators into high-resolution video generators.
Project Site: nvda.ws/3MVNUHn
Paper: nvda.ws/3Lk3uLS
12
71
319
98,356
Just released: 3D DL researchers can build on the latest algorithms to simplify and accelerate workflows using Kaolin PyTorch Library. Learn more: nvda.ws/3zKvgsD
71
326
This year at #NeurIPS2018, NVIDIA and @aivatechnology collaborated on a brand new #AI-composed song honoring legendary rock band Queen by training a deep learning-based system on hundreds of rock songs. nvda.ws/2BVK6AY
3
35
295
Explore one of the most critical components making “the world’s largest GPU” a reality - learn how NVSwitch integrates 16 GPUs to deliver NVIDIA #DGX-2: the world’s first 2 petaFLOPS #AI system. nvda.ws/2LeiloD
8
94
296
Capture, Render, Win
Announcing the #InstantNeRFVR Contest 🎉
Read more about how to create an immersive scene with our new #InstantNeRF VR software for a chance to win an NVIDIA GeForce RTX 3090 Ti graphics card.
➡️ nvda.ws/3Rk7opc
5
74
319
56,049
For the first time, @UCSF researchers trained and implemented a deep learning model on @NVIDIA #GPUs that can convert brain signals into speech. The work has the potential to one day help people who can't talk communicate. nvda.ws/2VBsTrk
8
108
299
Amazing! @UWaterloo researchers are using #AI + an @NVIDIA GPU to develop autonomous exoskeleton legs to help users walk, climb stairs, and avoid obstacles. Read how: nvda.ws/2PpqmPy
10
56
303
👀 @AIatMeta Llama 3.1 405B trained on 16K NVIDIA H100s - inference is #TensorRT #LLM optimized⚡
🦙 400 tok/s - per node
🦙 37 tok/s - per user
🦙 1 node inference
➡️ nvda.ws/3LB1iyQ✨
11
66
308
40,100
📣 Introducing FoundationPose, a unified foundation model for 6D object pose estimation and tracking for novel objects from #NVIDIAResearch.
See the details below.
🧵 👇
3
71
309
52,358
🖼️ Ready for next-level image generation?
@bfl_ml's FLUX.1 image generation model suite -- built on the Diffusion Transformer (DiT) architecture, and trained on 12 billion parameters -- is now accelerated by #TensorRT and runs the fastest ⚡️on NVIDIA RTX AI PCs.
🙌 Learn more ➡️ blogs.nvidia.com/blog/ai-dec…
#AIDecoded
12
40
302
55,223
🛠️ Fine-tune gpt-oss and deploy in FP4
Directly fine-tuning ultra-low precision models like gpt-oss in MXFP4 can potentially jeopardize model performance. In our latest blog, we share a reproducible fine-tuning + quantization-aware training (QAT) recipe that adapts gpt-oss to downstream tasks and restores the model to 4-bit precision for optimized inference deployment.
🔹 Step-by-step QAT breakdown with code examples
🔹 Benchmarks showing post fine-tuning + QAT task specific performance
🔹 How NVFP4 on NVIDIA Blackwell can push accuracy even further
The full recipe is available in the NVIDIA Model Optimizer repository so you can try it on your own
📖 Read the blog: nvda.ws/4p07Mts
8
56
311
16,928
🧠 AI reasoning is here: 3 key scaling laws are changing the world of #AI:
1️⃣ Pretraining scaling = bigger data, better performance
2️⃣ Post-training scaling = fine-tuning for precision
3️⃣ Test-time scaling (long thinking) = multi-pass reasoning for complex problems
Here’s what you need to know ➡️ nvda.ws/4hTI7hQ
45
80
295
53,658
Erik Lindahl from @Stockholm_Uni, John Stone from @Illinois_Alma, and Jim Phillips from @NCSAatIllinois join us at #SC18 to celebrate some of the major developer milestones in GPU-accelerated #HPC and #deeplearning history. nvda.ws/2K33nTB
2
37
262
So many innovative robotics on the show floor here at #GTC25 🤖
Check out this articulating hand for complex industrial tasks from @SCHUNKInc
14
59
305
14,095
Prototype, fine-tune, and deploy with ease. 🙌 NVIDIA DGX Spark, your own personal AI supercomputer.
✅ 1 petaFLOP
✅ 128GB of coherent, unified system memory
✅ ConnectX-7 Smart NIC
✅ Up to 4TB Storage
✅ 150mm L x 150mm W x 50.5mm H
Reserve yours now ➡️ nvda.ws/44IBhb1
21
49
302
28,166
Prototype, fine-tune, and deploy with ease 🙌 with NVIDIA DGX Spark, your own personal AI supercomputer.
✅ 1 petaFLOP
✅ 128GB of coherent, unified system memory
✅ ConnectX-7 Smart NIC
✅ Up to 4TB Storage
✅ 150mm L x 150mm W x 50.5mm H
⏰ Reserve yours now ➡️ nvda.ws/47iaC6A
Get ready to #SparkSomethingBig 💫
16
42
307
24,666
🎉 Congrats to our #NVIDIAResearch team for winning 🏆 1st Place and the 🏆 Innovation Award for the #CVPR2023 #3D Occupancy Challenge.
Their 3D object detection method substantially improved performance in comparison to prior approaches.
Read the paper: nvda.ws/3CoXTym
2
59
283
34,763
We recently open sourced the NVIDIA Nemotron Nano 9B v2 model (along with the data, and the technical report to recreate it).
We've also released the intermediate checkpoint, Nemotron Nano 12B v2, that was distilled into what would eventually become Nano 9B v2.
Read more here: nvda.ws/3HEsG0z
Check out the model on Hugging Face: nvda.ws/4mGhu2t
9
53
309
17,027
Watch how @UCBerkeley researchers used @NVIDIA #GPUs and reinforcement learning to train simulated characters to perform acrobatic moves by watching humans. nvda.ws/2OY1lc2
6
97
281
"By making @AIatMeta Llama models open source, we're committed to democratizing access to cutting-edge AI technology. With Project DIGITS, developers can harness the power of Llama locally, unlocking new possibilities for innovation and collaboration.”
-- @Ahmad_Al_Dahle, VP, GenAI at Meta
Announcing NVIDIA Project DIGITS, a personal AI supercomputer that’s powered by the NVIDIA GB10 Superchip and based on #NVIDIAGraceBlackwell architecture. nvda.ws/4fLyTm3
Preconfigured with the NVIDIA AI software stack, developers, researchers, data scientists and students can prototype, fine-tune and inference large AI models on their desktop and deploy them to the data center or cloud. #CES2025
26
49
285
108,877
🙌 7 Python drop-in replacements for massive GPU speedups—no code rewrites required:
pandas → %load_ext cudf.pandas
polars → .collect(engine="gpu")
scikit-learn → %load_ext cuml.accel
xgboost → device="cuda"
umap → %load_ext cuml.accel
hdbscan → %load_ext cuml.accel
networkx → %env NX_CUGRAPH_AUTOCONFIG=True
📝 Full blog with example demos + Colab notebooks: nvda.ws/3U5Cl2N
2
54
299
18,136
ICYMI: 👀📸 How NeRFs will help to re-imagine the world, an insightful blog by our community member @franclucent.
"The amount of applications are truly limitless, and I see it long term being the evolution of modern photography."
#InstantNeRF
➡️ bit.ly/3WgGIs8
1
60
285
44,232
A team of researchers from @UCSF just trained a #deeplearning system on @NVIDIA #GPUs to detect Alzheimer's disease years before an actual diagnosis. nvda.ws/2AVb5vY
6
79
277
We partnered with @huggingface to launch inference-as-a-service, which helps devs quickly prototype with open-source AI models hosted on the Hugging Face Hub and deploy them in production.
➡️nvda.ws/4bZ34EA✨
6
73
290
35,824
Microsoft today announced a breakthrough in conversational AI, training the largest transformer-based language generation model, based on over 17 billion parameters, using @NVIDIA DGX-2 systems.
3
79
277