NVIDIA AI(@NVIDIAAI)

WAMs and VLAs for Robot Learning | Cosmos Labs https://t.co/BcG6EQ7TUi

NVIDIA AI (@NVIDIAAI) — Thu, 13 Aug 2026 16:01:34 GMT

WAMs and VLAs for Robot Learning | Cosmos Labs x.com/i/broadcasts/1…

💬4🔄15❤️66👀10616📊15

Attn @cursor_ai devs 👀 Your agents can now tap into 300+ NVIDIA skills across 30+ products.

NVIDIA AI (@NVIDIAAI) — Thu, 13 Aug 2026 13:04:23 GMT

Attn @cursor_ai devs 👀

Your agents can now tap into 300+ NVIDIA skills across 30+ products.

Moshe Abramovitch@MosheAbram38005

NVIDIA skills are now in Cursor.

Add the @NVIDIA plugin and your agent picks up over 300 skills across 30+ products: CUDA, NeMo, RAG.

Tell it what you're building. It proposes the right skill and follows it.

Cursor MP: cursor.com/marketplace/nv…

Catalog: github.com/NVIDIA/skills

Your browser does not support the video tag.

💬32🔄46❤️731👀69704📊108

https://t.co/7M5cZb4uNb

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 22:09:38 GMT

x.com/thoughtworks/s…

Thoughtworks@thoughtworks

Congrats to @nvidia on the launch of Nemotron 3.5 Lightning. We got early access to test it and found it to be fast, highly customizable and controllable. Read Gustavo A. Lujan, Allen Roush and Andy Nolan's findings  

👉thoughtworks.com/insights/blog/…f

💬0🔄1❤️5👀4546📊1

https://t.co/2EigydTqJC

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 22:09:37 GMT

x.com/UnslothAI/stat…

Unsloth AI@UnslothAI

2-bit NVIDIA Nemotron 3.5 Lightning ran tool calls nonstop for 10 minutes on just 22GB of VRAM. 🤯

It cited 80+ websites, executed code & searched for 10 real-world locations.

Run and train via Unsloth Desktop.

GGUF: htthuggingface.co/unsloth/NVIDIA…ide: https://t.co/TbNdgO7qPf

Your browser does not support the video tag.

💬3🔄3❤️23👀8584📊4

Love seeing our partners post-training Nemotron 3.5 Lightning for their own domains, tools, and work...

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 22:09:25 GMT

Love seeing our partners post-training Nemotron 3.5 Lightning for their own domains, tools, and workflows, as well as those offering post-training support ⚡

Check out below to see what they’re building 🧵

💬20🔄42❤️356👀27450📊66

https://t.co/7M5cZb4uNb

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:49:22 GMT

x.com/thoughtworks/s…

Thoughtworks@thoughtworks

Congrats to @nvidia on the launch of Nemotron 3.5 Lightning. We got early access to test it and found it to be fast, highly customizable and controllable. Read Gustavo A. Lujan, Allen Roush and Andy Nolan's findings  

👉thoughtworks.com/insights/blog/…f

💬0🔄0❤️2👀436

https://t.co/KDHS7kioqT

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:49:13 GMT

x.com/tinkerapi/stat…

Tinker@tinkerapi

Nemotron 3.5 Lightning from @NVIDIAAI is out today and available on Tinker. With just 3B active parameters and optimized for throughput speed, 3.5 Lightning is designed for work where latency and cost matter.

💬0🔄0❤️1👀254

https://t.co/gkoZAC8geD

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:49:07 GMT

x.com/DeepCogito/sta…

Deep Cogito@DeepCogito

Excited to see @NVIDIAAI continue pushing the open model ecosystem forward with Nemotron 3.5 Lightning.

At @DeepCogito, we are big believers in open weight, customizable models and the role they’ll play in making frontier intelligence broadly accessible and useful.

NVIDIA has been a fantastic partner to startups like us building at the frontier, and we’re excited to be part of the NVIDIA post-training partner ecosystem.

Congrats to the @NVIDIAAI team on the launch!

💬0🔄0❤️1👀256

https://t.co/gtE1W9xX1b

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:49:01 GMT

x.com/togethercomput…

Together AI@togethercompute

NVIDIA Nemotron 3.5 Lightning is now live on Together AI.

The fastest open model in its class is built for always-on agents that need to complete high-volume, specialized work quickly.

💬0🔄0❤️1👀284

https://t.co/N92eiDZS6j

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:49 GMT

x.com/FireworksAI_HQ…

Fireworks@FireworksAI_HQ

Looking for a faster specialized model for your Agent Work?

@nvidia Nemotron 3.5 Lightning (30B MoE, 3B active params) is now live on Fireworks. It’s distilled from NVIDIA Nemotron 3 Ultra to be your high-volume agent engine.

With strong performance on PinchBench and top-tier scores in AA-Omniscience Non-Hallucination, Nemotron 3.5 Lightning is engineered for reliable, high-productivity agentic workflows.

Built for specialization, not generalization. Start building:

app.fireworks.ai/models/firewor…

💬0🔄0❤️3👀265

https://t.co/VrKrN6muut

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:39 GMT

x.com/LilaSciences/s…

Lila Sciences@LilaSciences

We're building scientific superintelligence with @nvidia's Nemotron 3.5 Lightning. Here's more on what this means for improving reasoning capabilities: blogs.nvidia.com/blog/nemotron-…

💬0🔄0❤️0👀89

https://t.co/pDrxCmnxHe

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:34 GMT

x.com/baseten/status…

Baseten@baseten

NVIDIA Nemotron 3.5 Lightning is live on Baseten day 0!

This is the fastest open model in its class, built for long-running agents. Compared to similar-sized open models, it offers:

- 4x higher throughput
- 50% lower cost 
- 63.4% fewer output tokens in production testing
- 30B MoE architecture (3B active)
- 1M token context length

💬0🔄0❤️2👀74

https://t.co/JYcidGNmvS

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:27 GMT

x.com/appliedcompute…

Applied Compute@appliedcompute

Nemotron 3.5 Lightning by @NVIDIAAI is now supported for training and inference on the Applied Compute Platform.

On our agentic coding benchmark, decode throughput, time to first token, and median user latency remained effectively unchanged as concurrency and total token throughput scaled 16x.

Its LatentMoE and Mamba architecture lets us scale sparsity, context length, and batch size with minimal overhead, dramatically increasing iteration throughput across post-training runs.

💬0🔄0❤️1👀88

https://t.co/C7QpaGkz3X

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:22 GMT

x.com/locai_labs/sta…

Locai Labs@locai_labs

x.com/i/article/2087…

💬0🔄0❤️0👀69

https://t.co/wzwzSyMKyL

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:12 GMT

x.com/DreamGroupAI/s…

Dream Security@DreamGroupAI

Today, @nvidia launched Nemotron 3.5 , its new open model built for fast, high-volume execution in agentic AI systems.

Dream had early access to the model. Our researchers, led by Guy Feigenblat, Shai Nahum Gefen and Dmitry Basin, invested extensively in supervised fine-tuning, post-training, domain adaptation, cybersecurity reasoning and agentic tool use to advance Dream’s proprietary agentic cybersecurity conversational model.

Pleased to see strong performance across Dream’s internal cybersecurity benchmarks that will help our customers better understand their security posture, investigate risks and turn large volumes of security data into clear, actionable insights. 

Full launch blog here:
developer.nvidia.com/blog/nvidia-ne…

💬0🔄0❤️1👀84

https://t.co/hOo42SeZa6

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:06 GMT

x.com/distil_labs/st…

distil labs@distil_labs

Nemotron 3.5 Lightning from @NVIDIAAI is out. As one of a handful of European launch partners, we fine-tuned it against 4 comparable MoE models on 6 tasks.

Prompted, third. Fine-tuned, first.

Full numbers, and where it loses, from @j_golebiowski: distillabs.ai/blog/the-best-…

Your browser does not support the video tag.

💬0🔄0❤️0👀104

https://t.co/WLyKumh5ch

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:48:01 GMT

x.com/trajectorylabs…

Trajectory@trajectorylabs

Continual learning is a bet that the retraining loop will get cheaper over time. With larger models, you can maybe run this loop once every few weeks. But with smaller models, you can run it nightly, per customer. And it keeps recursing: a model per company, then a model per client that company serves, then per matter. We’re getting closer to intelligence cheap enough to meter.

On the path to this, we received early access to, and post-trained @nvidia's Nemotron 3.5 Lightning on @harvey LAB. One click on the Trajectory platform, no new engineering. 0% to 8.3%, above Opus 4.6 at 6.6%.

💬0🔄0❤️1👀250

https://t.co/EqHNOzY2q7

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:46 GMT

x.com/AgileRL_Inc/st…

AgileRL@AgileRL_Inc

We are proud to announce a collaboration between AgileRL and @NVIDIAAI  to support the post-training of Nemotron models. NVIDIA's Nemotron open-source model family is now available for fine-tuning on Arena, our platform for creating AI agents specialized at any task.

NVIDIA gave us early access to their latest 30B3A parameter MoE model, Nemotron 3.5 Lightning. On Arena, we trained it to outperform Claude Sonnet 5 on two different tasks: a long-horizon reasoning challenge and a real customer support workload. Nemotron proved exceptionally easy to post-train: a single six-hour run on four H100 GPUs was enough to master the reasoning task, at context lengths beyond 50,000 tokens. The full results are in the announcement, linked below.

With this collaboration, Arena customers get access to NVIDIA Nemotron models as soon as they are released, with the training and evaluation setup already built around them. Post-train Nemotron on your own data, environment and edge cases, to build an agent that masters your task.

Across industries including banking, insurance, aerospace and government, production traffic is moving off frontier model APIs and onto infrastructure these companies control. Businesses want agents with genuine expertise in their specific task, trained on their own data. Security, control and data sovereignty demand that model weights remain on their own infrastructure. And they want the fixed cost of hardware they own, rather than per-token pricing that compounds with every request.

This collaboration provides that path. Begin with a dataset, an RL environment, or simply a description of the job the agent has to do, and our team will build the rest with you. Arena handles the training and deploys the finished agent in one click, with the weights yours to keep.

💬0🔄0❤️1👀315

https://t.co/FqhQKeutuS

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:41 GMT

x.com/ReasonableIO/s…

Reasonable@ReasonableIO

Smaller, faster, capable: writing machine-checked proofs with a 30B open-weight model

We fine-tuned NVIDIA’s latest open model Nemotron 3.5 Lightning on 4B tokens of synthetic Verus data. The result: it beats a model ~50x its size on per-attempt pass rate while nearly matching it on pass@3, with faster token generation than any similarly sized open-weight model we tested.

Explore how the models we tested fail at formal proofs, how they cheat, and the effects of fine-tuning: 
reasonable.io/blog/verificat…

💬1🔄0❤️1👀514📊1

https://t.co/zLd1hzAsJB

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:22 GMT

x.com/uniphore/statu…

Uniphore@uniphore

We tested @NVIDIAAI's Nemotron 3.5 Lightning against real enterprise agentic workloads & found: 5x throughput vs Gemma 4 31B IT at matched parameter count. Accuracy gains moved Nemotron 3.5 Lightning into evaluation for high-volume paths in the Uniphore's Business AI Cloud

💬0🔄0❤️0👀128

https://t.co/H0II4NxEnz

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:16 GMT

x.com/harvey/status/…

Harvey@harvey

We post-trained @NVIDIAAI Nemotron 3.5 Lightning on Legal Agent Bench with @trajectorylabs.

Here's what we found:

1) Post-training improved agent performance from 0% to 8.3% on held-out LAB tasks, beating both Opus 4.6 and the much larger post-trained Nemotron 3 Ultra.

2) Performance improved across nine practice areas with no regressions.

3) Post-training reduced average model output from 90k to 37k tokens, increasing the model's reward-per-token by 2.4x.

Through our collaboration with NVIDIA and Trajectory we’re committed to pushing the frontier of legal intelligence and cost efficiency with open weight models.

Deep dive:

💬0🔄0❤️1👀92

https://t.co/2EigydTqJC

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:10 GMT

x.com/UnslothAI/stat…

Unsloth AI@UnslothAI

2-bit NVIDIA Nemotron 3.5 Lightning ran tool calls nonstop for 10 minutes on just 22GB of VRAM. 🤯

It cited 80+ websites, executed code & searched for 10 real-world locations.

Run and train via Unsloth Desktop.

GGUF: htthuggingface.co/unsloth/NVIDIA…ide: https://t.co/TbNdgO7qPf

Your browser does not support the video tag.

💬0🔄0❤️2👀114

https://t.co/r7p9CKh4w8

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:47:03 GMT

x.com/fastinoAI/stat…

Fastino Labs@fastinoAI

In collaboration with @nvidia we're releasing two new open weight models: Fastino-Nemotron-3.5-Lightning-Finance and Fastino-Nemotron-3.5-Lightning-Healthcare.

Working closely with the Nemotron team, we developed both models on Nemotron 3.5 Lightning using the Fastino Fine-Tuning Agent, which autonomously ran the entire post-training pipeline. The base model was highly responsive, with the agent reaching substantial gains across both domains with compact datasets alone.

- Fastino-Nemotron-3.5-Lightning-Finance excels at numerical reasoning, research, and summarization over financial concepts and documents. When evaluated, the fine-tuned model increased FinQA execution accuracy by 43.37 points (15.86% to 59.23%), and increased 7.81 points (from 49.65% to 57.46%) on BizFinBench.

- Fastino-Nemotron-3.5-Lightning-Healthcare specializes in clinical conversation quality, summarizing notes and discharge documentation, and extracting medical concepts from unstructured text. In post-training, the Fastino Fine-Tuning Agent improved performance by a margin of 4.28 points on HealthAdminBench (from 25.67% to 29.95%) and 5.09 points on MedCalc-Bench (from 49.09% to 54.18%). It also increased flag accuracy on MEDEC by 11.32 points, beating Opus 4.6 and beating Muse Spark by 0.98 points.

Both models are available on @huggingface under the Apache 2.0 license:
- huggingface.co/fastino/Fastin…
- huggingface.co/fastino/Fastin…

Today we are also releasing the Fastino Fine-Tuning Agent in private preview, which you can sign up for access to here: fastino.ai/nvidia-collabo…

💬0🔄0❤️1👀90

https://t.co/bbQ3Pdi8Du

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:46:41 GMT

x.com/coderabbitai/s…

CodeRabbit@coderabbitai

Easy to train. Smart at routing. Efficient at scale.

In collaboration with Baseten, we fine-tuned NVIDIA Nemotron 3.5 Lightning on CodeRabbit's routing decisions in under three hours and for less than $100.

The result?

> ~4% higher accuracy than our previous GPT-class model.
> ~50% lower inference cost.

Full breakdown:
coderabbit.ai/blog/teaching-…

💬0🔄0❤️1👀154

https://t.co/fgCiHVwmcI

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:46:32 GMT

x.com/CrowdStrike/st…

CrowdStrike@CrowdStrike

We’re excited to see the latest NVIDIA Nemotron model and continue pushing the boundaries of what AI can do for cybersecurity. ⚡️

CrowdStrike customized Nemotron 3.5 Lightning for cybersecurity agent workflows, achieving analyst-grade accuracy with faster training and significantly lower compute requirements.

The result: helping auto-close up to 88% of benign detections, so security teams can focus on the threats that matter most.

💬0🔄0❤️1👀102

Love seeing our partners post-training Nemotron 3.5 Lightning for their own domains, tools, and work...

NVIDIA AI (@NVIDIAAI) — Wed, 12 Aug 2026 21:46:02 GMT

Love seeing our partners post-training Nemotron 3.5 Lightning for their own domains, tools, and workflows, as well as those offering post-training support ⚡

Check out below to see what they’re building 🧵

💬23🔄3❤️19👀1780📊21

@huggingface And it ships with its data. Alongside Lightning, we’re releasing Nemotron-RL-Agentic-T...

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 19:23:54 GMT

@huggingface And it ships with its data.

Alongside Lightning, we’re releasing Nemotron-RL-Agentic-Terminal-Pivot, the open agentic RL dataset used to post-train its coding agent capabilities. 

Dataset on @huggingface 🤗→huggingface.co/datasets/nvidi…7

💬3🔄7❤️75👀13021📊14

Run Open Models Locally: Nemotron 3 Ultra on DGX Station | Nemotron Labs https://t.co/fOHrFGpOCl

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 18:01:04 GMT

Run Open Models Locally: Nemotron 3 Ultra on DGX Station | Nemotron Labs x.com/i/broadcasts/1…

💬10🔄18❤️105👀10451📊23

Learn Robot Manipulation with LeRobot and ROS 2 | NVIDIA Jetson AI Lab https://t.co/RiaYWP39eg

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 16:00:21 GMT

Learn Robot Manipulation with LeRobot and ROS 2 | NVIDIA Jetson AI Lab x.com/i/broadcasts/1…

💬14🔄15❤️93👀11001📊26

As always, NVIDIA Nemotron 3.5 Lightning is open and customizable. This includes weights, data and ...

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 13:46:10 GMT

As always, NVIDIA Nemotron 3.5 Lightning is open and customizable.

This includes weights, data and recipes.
Available now on @huggingface 🤗 → huggingface.co/nvidia/NVIDIA-…Z

💬18🔄22❤️301👀35084📊59

Not every step in an agent workflow needs the same model. That’s why we’re also releasing NVIDIA Ne...

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 13:44:55 GMT

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

That’s why we’re also releasing NVIDIA NeMo Switchyard, a new open source library for model routing.

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

Learn more: nvda.ws/457eK7l

Your browser does not support the video tag.

💬4🔄5❤️93👀14073📊16

Long-running agents spend most of their time executing: calling tools, validating results and delega...

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 13:37:50 GMT

Long-running agents spend most of their time executing: calling tools, validating results and delegating work.

Nemotron 3.5 Lightning is built for this high-volume execution, at a size that can run anywhere from an NVIDIA DGX Spark to the data center.

See it running agentic workflows on DGX Spark and read the technical breakdown: nvda.ws/4heK23m

Your browser does not support the video tag.

💬3🔄1❤️7👀946📊4

Introducing NVIDIA Nemotron 3.5 Lightning⚡ An open 30B MoE model with 3B active parameters, built f...

NVIDIA AI (@NVIDIAAI) — Tue, 11 Aug 2026 13:00:09 GMT

Introducing NVIDIA Nemotron 3.5 Lightning⚡

An open 30B MoE model with 3B active parameters, built for always-on agents to complete high-volume, specialized tasks faster.

It delivers up to 4x the output speed of similar-sized models.

💬317🔄649❤️5641👀2361530📊1224

@AIatMeta https://t.co/eebwQpQMCc

NVIDIA AI (@NVIDIAAI) — Mon, 10 Aug 2026 13:30:13 GMT

@AIatMetax.com/sgl_project/st…

SGLang@sgl_project

SGLang is honored to provide day-0 support for @AIatMeta's Muse Glimmer.

~230 tok/s on a single RTX 5090 with NVFP4 + DFlash on, and it runs out of the box on @NVIDIAAI RTX PRO 6000, DGX Spark, and Apple Silicon via MLX.

Huge thanks to the NVIDIA and Meta teams for the collaboration. Go build your local agent 🚀

💬0🔄1❤️22👀5839📊2

@AIatMeta https://t.co/VPWL6oRtBq

NVIDIA AI (@NVIDIAAI) — Mon, 10 Aug 2026 13:30:05 GMT

@AIatMetax.com/vllm_project/s…

vLLM@vllm_project

@Meta is back in open source.

Excited to announce Day-0 vLLM support for Muse Glimmer 30B, the first open-weights model from Meta Superintelligence Labs — which ships under Apache 2.0!!!

30B dense, 128K+ context, multimodal, built for local agents.

Capable enough for long-horizon tasks, small enough to run on hardware you own.

Try it now on your device:
vllm serve meta-models/Muse-Glimmer-30B

Kudos to @inferact, @AIatMeta, and @NVIDIAAI for bringing the model alive in vLLM!

💬0🔄0❤️9👀1851📊1

Great to see @AIatMeta back publishing open models 🙌 Muse Glimmer is a 30B open-weight dense model...

NVIDIA AI (@NVIDIAAI) — Mon, 10 Aug 2026 13:29:29 GMT

Great to see @AIatMeta back publishing open models 🙌

Muse Glimmer is a 30B open-weight dense model with a 120K+ context window, built for long-running agents, delivering up to 20K tokens/sec on a single GPU.

It’s optimized to run locally across NVIDIA edge, desktop, and workstation AI platforms.

Try it with our GPU-accelerated endpoint:nvda.ws/45ehAYeL

AI at Meta@AIatMeta

Introducing Muse Glimmer, an open-weight 30B-parameter model optimized for local, always-on agent workflows.

Muse Glimmer delivers strong performance on key agentic use cases and benchmarks compared with leading models in its size category, and is designed to run entirely on consumer hardware like a Mac or PCs with performant GPUs.

In keeping with our long tradition of sharing fundamental AI research, we’re releasing model weights under a permissive Apache 2.0 license.

🧵👇

💬47🔄76❤️1043👀65328📊126

@MiaAI_lab https://t.co/zJU4GuHSVq

NVIDIA AI (@NVIDIAAI) — Wed, 05 Aug 2026 19:38:07 GMT

@MiaAI_labx.com/alecqfong/stat…

Alec Fong@alecqfong

Current local ai stack 

2x DGX Sparks - DeepSeek v4 flash (90 tok/sec) - implementation & sub agents

2x DGX Stations - GLM 5.2 nvfp4 (120 tok/sec) - planning and advising

Brev for cluster networking, always on cloud agent, and orchestration

💬1🔄0❤️16👀5919📊2

@MiaAI_lab https://t.co/1eIWpH9Vx7

NVIDIA AI (@NVIDIAAI) — Wed, 05 Aug 2026 19:36:06 GMT

@MiaAI_labx.com/takuz0_/status…

Takuya KAWATSU@takuz0_

Testing 8x DGX Spark nodes in an open world setup :)

💬2🔄1❤️23👀7400📊4

The time for local AI is here. So many great models have been released recently, here's a guide fr...

NVIDIA AI (@NVIDIAAI) — Wed, 05 Aug 2026 19:35:09 GMT

The time for local AI is here. 

So many great models have been released recently, here's a guide from @MiaAI_lab on chaining DGX Sparks together to run them.

Drop a pic of your setup and let us know what you're running 👇

Mia@MiaAI_lab

What are the best models you can run on your @NVIDIAAI DGX Spark? ✨

Aug 2026 Edition

1× DGX Spark
• DeepSeek v4 Flash - 1M ctx, 26 tok/s - recommended!
• ⁠Qwen 3.6 35b NVFP4 - 256k ctx, 81 tok/s
• ⁠Qwen 3.6 27b NVFP4 - 256k ctx, 33 tok/s

⁠Qwen 3.8 27b - should be released soon and this could change my recommendation!

2× DGX Sparks ← sweet spot!
• DeepSeek v4 Flash 0731 - 1M ctx, 82 tok/s
• Inkling-Small - 1M context, Full Omni, 33 tok/s
• MiMo-V2.5 - 1M ctx, Full Omni, 31 tok/s
• Step-3.7-Flash — 256K ctx, 30 tok/s

3× DGX Sparks
• GLM-5.2 with Vision - 348k context, 25 tok/s - still the best intelligence you can run locally if you have 3 sparks.
• DeepSeek v4 Flash on 2 units + smaller models on the 3rd spark for images, ComfyUI and other things. This setup gives you DeepSeek v4 Flash speeds for coding, plus strong agentic workflows and image support from smaller models.

4× DGX Sparks
• GLM 5.2 NVFP4 across all 4 units - still think this is the way to go if you have 4 units!
• The alternative is running DeepSeek v4 Flash on 2 units and any other 2x setup on the other units. The choice is your.

Links and repos below 👇

💬63🔄54❤️800👀64055📊133

Dassault Systèmes Accelerates Simulation with NVIDIA AI & GPUs https://t.co/2uRhCslgJ5

NVIDIA AI (@NVIDIAAI) — Wed, 05 Aug 2026 16:05:02 GMT

Dassault Systèmes Accelerates Simulation with NVIDIA AI & GPUs https://t.co/2uRhCslgJ5

💬9🔄7❤️72👀12127📊18

https://t.co/dD2JgpM8Xj

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 23:31:25 GMT

x.com/MistralAI/stat…

Mistral AI@MistralAI

🛡️Introducing Shieldstral, Mistral’s 3B open-weights model for content safety that can be deployed on-device �mistral.ai/news/shieldstr…9p

💬2🔄2❤️9👀5737📊3

https://t.co/OUJb2clkVl

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 19:24:41 GMT

x.com/databricks/sta…

Databricks@databricks

Databricks joins the Open Secure AI Alliance to advance AI safety and security, alongside @nvidia and other industry leaders. 

The alliance is built on a simple but powerful premise: AI safety and security research should be shared openly, and the tools it produces should be built on open systems. 

Databricks brings governed data, model and agent capabilities, and open tools and frameworks for security, governance, red teaming, and cyber defense.

Read more about what Databricks is bringing to the alliance and how we're securing the agentic era together databricks.com/blog/databrick…

💬2🔄2❤️15👀6487📊3

https://t.co/aBfiTJnQ9U

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 18:38:22 GMT

x.com/CrowdStrike/st…

CrowdStrike@CrowdStrike

Cybersecurity is strongest when we work together.

As a founding member of the Open Secure AI Alliance, we're proud to help shape the proposed SAFE Working Group alongside @Cisco, @huggingface, @NVIDIA, @RedHat and others.

Learn more: crwdstr.ke/6012BEh6Cm

💬1🔄1❤️7👀4469📊2

How a Chess Grandmaster Built an Autonomous Business | Nemotron Labs https://t.co/vdzoTvY54e

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 18:00:09 GMT

How a Chess Grandmaster Built an Autonomous Business | Nemotron Labs x.com/i/broadcasts/1…

💬4🔄4❤️67👀11542📊12

https://t.co/3W6tB8lS07

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 17:37:34 GMT

x.com/uipath/status/…

UiPath@UiPath

More eyes on the problem, more hands building the solution. We’re joining the Open Secure AI Alliance to help develop open tools that strengthen security for software and AI agents. 🤝

💬1🔄0❤️1👀626📊1

https://t.co/ALoQkpIqa0

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 17:37:34 GMT

x.com/fortanix/statu…

Fortanix@fortanix

We're excited to see Fortanix join the @nvidia Open Secure AI Alliance, an important industry initiative bringing together technology leaders to advance AI safety & security through open collaboration.

httphubs.li/Q04rQ5Sl0ec#secureain#confidentialais#aisecuritytasecurity

💬1🔄0❤️4👀4096📊1

120+ members and counting! Here’s what they’re saying 🧵

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 17:37:32 GMT

120+ members and counting!

Here’s what they’re saying 🧵

NVIDIA@nvidia

Security is strongest when organizations collaborate openly and move quickly. 

Now over 120 members strong, the Open Secure AI Alliance is sharing new open source security contributions, including new proposed SAFE guidelines, for turning confidential incident findings into stronger protection across the ecosystem.

💬6🔄7❤️84👀19367📊13

Build NemoClaw Agents on Jetson | NVIDIA Jetson AI Lab https://t.co/JUSahmfEG4

NVIDIA AI (@NVIDIAAI) — Tue, 04 Aug 2026 16:00:08 GMT

Build NemoClaw Agents on Jetson | NVIDIA Jetson AI Lab x.com/i/broadcasts/1…

💬6🔄13❤️113👀11460📊22