Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Video: Decision model trained in ~2 mins on a single DGX Spark using Unsloth! Qwen3.5-2B as base model. Just follow instruction here:

Ivan Fioravanti

44,347 subscribers

24,226 views • 20 days ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

16 Comments

Ivan Fioravanti20 days ago

On the left sparkDash v2 by @MiaAI_lab that is something incredible to monitor your DGX Sparks! Link:

Ivan Fioravanti20 days ago

I need to train it more.

Daniel Han20 days ago

Nice!! Will fix the model slow loading issue haha

Ivan Fioravanti20 days ago

Yeah I noticed it, but after that it's faster than light!

WestWaters20 days ago

Nice just dropped a clef flash Pollard version from @Cloudflare if interested🫡

Maziyar PANAHI20 days ago

@UnslothAI beautiful! let’s put it up and evaluate it too 😆

AriesTheCoder20 days ago

Thank you bro.

Suraj Verma20 days ago

Unsloth's Qwen3.5-2B benchmark reports around 40 minutes of training for 81% test accuracy. Curious how many steps this 2-minute DGX Spark run completed and what accuracy it reached on a held-out test set. That would make the speed comparison much more useful.

Mike Gannotti 🔳20 days ago

Thank for the share!

Ivan Fioravanti20 days ago

🫡

Jonas Lomholdt20 days ago

@UnslothAI Could it work on a 48gb MacBook pro?

Ivan Fioravanti20 days ago

@UnslothAI Yes!

Eugene Pyvovarov20 days ago

interesting, at first it took it a minute to load? seems liek so small model

Ivan Fioravanti20 days ago

@danielhanchen is aware and working on this 💪

John Kovacs20 days ago

unsloth getting a decision model done in about 2 mins on one dgx spark is the bar i'd steal. my last local fine-tune still ate an afternoon

Tablet1 User120 days ago

The last question though... it failed badly!

Related Videos

I just trained my own Jev-style Decision model locally! With Qwen3.5 0.8B as the base, accuracy jumped from 37% to 65% in just 60 steps (10 minutes of training). All on just 4GB of VRAM. This video shows how you can do the same. Chapters: 0:00 - Intro 0:50 - Set up Unsloth Studio 3:47 - Pick a model and dataset 5:19 - Train with LoRA 7:51 - Test it in the app 10:03 - Same workflow in code 12:38 - Train and evaluate (37% → 65%) 14:46 - Save and run inference 15:24 - Outro If you’re new to JEV and want to understand how decision models work, I’ve written a detailed article about it. It’s quoted below.

I just trained my own Jev-style Decision model locally! With Qwen3.5 0.8B as the base, accuracy jumped from 37% to 65% in just 60 steps (10 minutes of training). All on just 4GB of VRAM. This video shows how you can do the same. Chapters: 0:00 - Intro 0:50 - Set up Unsloth Studio 3:47 - Pick a model and dataset 5:19 - Train with LoRA 7:51 - Test it in the app 10:03 - Same workflow in code 12:38 - Train and evaluate (37% → 65%) 14:46 - Save and run inference 15:24 - Outro If you’re new to JEV and want to understand how decision models work, I’ve written a detailed article about it. It’s quoted below.

Akshay 🚀

58,235 views • 21 days ago

Qwen3.8 Flash on a single DGX Spark. Smooth experience!

Qwen3.8 Flash on a single DGX Spark. Smooth experience!

Mia

18,300 views • 1 month ago

Receipts in video, see it float at ~100-150 while coding the fluctuations were for task and context switching of the model. This thing rips through code! A Single NVIDIA AI DGX Spark ⚡️

Receipts in video, see it float at ~100-150 while coding the fluctuations were for task and context switching of the model. This thing rips through code! A Single NVIDIA AI DGX Spark ⚡️

ÆON FORGE ✨

10,822 views • 4 months ago

🚨 NEW ON AND make NVIDIA's Nemotron 3.5 Lightning faster. Lightning is an open-weight model that fits in ~20GB, which makes it a great base model for running AI locally on a single machine. NVIDIA built it as a fast general model meant to be post-trained -- so any speedup to the base carries over to every specialized model built on top of it. On day 1, solvers on Yukon already got it running 31.9% faster on DGX Spark and 13.8% faster on Mac. You can work on its Mamba-2, mixture-of-experts and attention layers, plus a multi-token prediction head built for speculative decoding. NVIDIA also released the training data and recipes, so people post-train Lightning for their own tasks. Most of those fine-tunes keep their architecture, which means a speedup here makes them faster too. Two tracks: • Apple Silicon → • NVIDIA DGX Spark → The challenge runs until Oct 29, 2026 To join, pick a track, point your agent at it and let it rip.

🚨 NEW ON AND make NVIDIA's Nemotron 3.5 Lightning faster. Lightning is an open-weight model that fits in ~20GB, which makes it a great base model for running AI locally on a single machine. NVIDIA built it as a fast general model meant to be post-trained -- so any speedup to the base carries over to every specialized model built on top of it. On day 1, solvers on Yukon already got it running 31.9% faster on DGX Spark and 13.8% faster on Mac. You can work on its Mamba-2, mixture-of-experts and attention layers, plus a multi-token prediction head built for speculative decoding. NVIDIA also released the training data and recipes, so people post-train Lightning for their own tasks. Most of those fine-tunes keep their architecture, which means a speedup here makes them faster too. Two tracks: • Apple Silicon → • NVIDIA DGX Spark → The challenge runs until Oct 29, 2026 To join, pick a track, point your agent at it and let it rip.

Kydo

17,287 views • 7 days ago

How to Fine Tune Llama 3 for Better Instruction Following? 🚀 Setup Environment 📊 Train Model 💾 Save Model Locally 🌐 Upload Model to Hugging Face 📈 Evaluate Performance 🛠 Using 🦥 Unsloth AI Subscribe: YT: Meta AI at Meta Daniel Han

How to Fine Tune Llama 3 for Better Instruction Following? 🚀 Setup Environment 📊 Train Model 💾 Save Model Locally 🌐 Upload Model to Hugging Face 📈 Evaluate Performance 🛠 Using 🦥 Unsloth AI Subscribe: YT: Meta AI at Meta Daniel Han

Mervin Praison

14,111 views • 2 years ago

Mark Zuckerberg introducing Meta's Muse — "Muse spark model is trained specifically to excel as a personal agent."

Mark Zuckerberg introducing Meta's Muse — "Muse spark model is trained specifically to excel as a personal agent."

Yahoo Finance

65,802 views • 1 month ago

Ex-NVIDIA engineer who built Unsloth explained RL, kernels, reasoning, quantization, and agents in 2 hours 42 minutes - better than $5000 fine-tuning bootcamps. pick the base model -> write triton kernels for 2x faster fine-tune -> quantize to 4-bit -> run GRPO/DPO -> ship a reasoning model on your single GPU. That loop is why Unsloth is the default way to fine-tune Llama, Qwen, Gemma, and Phi on hardware you already own. Unsloth + Triton kernels + 4-bit quantization + GRPO/DPO + single-GPU fine-tuning - that's the stack. Watch and save it, then fine-tune your first model tonight.

Ex-NVIDIA engineer who built Unsloth explained RL, kernels, reasoning, quantization, and agents in 2 hours 42 minutes - better than $5000 fine-tuning bootcamps. pick the base model -> write triton kernels for 2x faster fine-tune -> quantize to 4-bit -> run GRPO/DPO -> ship a reasoning model on your single GPU. That loop is why Unsloth is the default way to fine-tune Llama, Qwen, Gemma, and Phi on hardware you already own. Unsloth + Triton kernels + 4-bit quantization + GRPO/DPO + single-GPU fine-tuning - that's the stack. Watch and save it, then fine-tune your first model tonight.

h100envy

497,788 views • 3 months ago

Qwen3.5-2B vs Qwen3.5-9B rap battle running locally on 2 foidable phones

Qwen3.5-2B vs Qwen3.5-9B rap battle running locally on 2 foidable phones

alexine 🏴‍☠️

134,796 views • 8 months ago

🔥 We fine-tuned Gemma 3 using Unsloth AI on a custom instruction dataset scraped and built with . The model is now able to answer questions about OpenAI Agents SDK documentation. Check out the demo below👇

🔥 We fine-tuned Gemma 3 using Unsloth AI on a custom instruction dataset scraped and built with . The model is now able to answer questions about OpenAI Agents SDK documentation. Check out the demo below👇

Eric Ciarla (hiring)

28,794 views • 1 year ago

Fine-tuning in 2026 has never been easier You can make any open-source model 10x more powerful And thanks to Unsloth Studio, creating custom datasets takes just a few mins, Here is the full course:

Fine-tuning in 2026 has never been easier You can make any open-source model 10x more powerful And thanks to Unsloth Studio, creating custom datasets takes just a few mins, Here is the full course:

David Ondrej

36,391 views • 5 months ago

This afternoon I picked up a new Nvidia DGX Spark computer with the goal of trying to run Gemma 4 31b (4bit) on it locally as a server. Just 1.5 hours later, it’s working! Using Open WebUI on my MacBook as the interface and it’s connecting to my DGX Spark running as a Gemma 4 server.

This afternoon I picked up a new Nvidia DGX Spark computer with the goal of trying to run Gemma 4 31b (4bit) on it locally as a server. Just 1.5 hours later, it’s working! Using Open WebUI on my MacBook as the interface and it’s connecting to my DGX Spark running as a Gemma 4 server.

Dave Lee

141,105 views • 6 months ago

OpenAI shows how gpt-oss can autonomously beat 2048 using reinforcement learning (RL). Training was done locally with Unsloth on NVIDIA DGX Spark. You can also do it free on Colab. 🦥 OpenAI DevDay notebook:

OpenAI shows how gpt-oss can autonomously beat 2048 using reinforcement learning (RL). Training was done locally with Unsloth on NVIDIA DGX Spark. You can also do it free on Colab. 🦥 OpenAI DevDay notebook:

Unsloth AI

98,958 views • 1 year ago

We made it possible to train your own Decision model locally on just 3GB VRAM! We converted Qwen, Gemma, Llama all into decision models by fine-tuning using a Clef head, boosting accuracy from 30 to up to 78%. You can try it yourself via Unsloth Desktop!

We made it possible to train your own Decision model locally on just 3GB VRAM! We converted Qwen, Gemma, Llama all into decision models by fine-tuning using a Clef head, boosting accuracy from 30 to up to 78%. You can try it yourself via Unsloth Desktop!

Daniel Han

27,256 views • 23 days ago

GLM 5.2 @ 19.8 tok/sec on 2 DGX Sparks ⚡️ I switched to a dspark draft model with K=2 (model from Red Hat AI) Acceptance is ~68% because the drafter was trained with the full FP8 model... Next step is to fine tune the drafter so acceptance on this super quant goes up

GLM 5.2 @ 19.8 tok/sec on 2 DGX Sparks ⚡️ I switched to a dspark draft model with K=2 (model from Red Hat AI) Acceptance is ~68% because the drafter was trained with the full FP8 model... Next step is to fine tune the drafter so acceptance on this super quant goes up

Joe Muller

73,297 views • 3 months ago

Running Qwen3.5-397B-A17B (4bit quants, 177 GB) on two DGX Sparks using llama.cpp with RPC and RDMA:

Running Qwen3.5-397B-A17B (4bit quants, 177 GB) on two DGX Sparks using llama.cpp with RPC and RDMA:

Radoslav Gerganov

26,018 views • 6 months ago

Unsloth AI and NVIDIA are Revolutionizing Local LLM Fine-Tuning: From RTX Desktops to DGX Spark Fine-tune popular AI models faster with Unsloth on NVIDIA RTX AI PCs such as GeForce RTX desktops and laptops to RTX PRO workstations and the new DGX Spark to build personalized assistants for coding, creative work, and complex agentic workflows. The landscape of modern AI is shifting. We are moving away from a total reliance on massive, generalized cloud models and entering the era of local, agentic AI. Whether it is tuning a chatbot to handle hyper-specific product support or building a personal assistant that manages intricate schedules, the potential for generative AI on local hardware is boundless. However, developers face a persistent bottleneck: How do you get a Small Language Model (SLM) to punch above its weight class and respond with high accuracy for specialized tasks? The answer is Fine-Tuning, and the tool of choice is Unsloth. Unsloth provides an easy and high-speed method to customize models. Optimized for efficient, low-memory training on NVIDIA GPUs, Unsloth scales effortlessly from GeForce RTX desktops and laptop all the way to the DGX Spark, the world’s smallest AI supercomputer...... Full analysis: NVIDIA NVIDIA AI NVIDIA AIDev NVIDIAnewsroom Unsloth AI Unsloth

Unsloth AI and NVIDIA are Revolutionizing Local LLM Fine-Tuning: From RTX Desktops to DGX Spark Fine-tune popular AI models faster with Unsloth on NVIDIA RTX AI PCs such as GeForce RTX desktops and laptops to RTX PRO workstations and the new DGX Spark to build personalized assistants for coding, creative work, and complex agentic workflows. The landscape of modern AI is shifting. We are moving away from a total reliance on massive, generalized cloud models and entering the era of local, agentic AI. Whether it is tuning a chatbot to handle hyper-specific product support or building a personal assistant that manages intricate schedules, the potential for generative AI on local hardware is boundless. However, developers face a persistent bottleneck: How do you get a Small Language Model (SLM) to punch above its weight class and respond with high accuracy for specialized tasks? The answer is Fine-Tuning, and the tool of choice is Unsloth. Unsloth provides an easy and high-speed method to customize models. Optimized for efficient, low-memory training on NVIDIA GPUs, Unsloth scales effortlessly from GeForce RTX desktops and laptop all the way to the DGX Spark, the world’s smallest AI supercomputer...... Full analysis: NVIDIA NVIDIA AI NVIDIA AIDev NVIDIAnewsroom Unsloth AI Unsloth

Marktechpost AI Dev News ⚡

31,551 views • 10 months ago

2.3x faster. Ran Unsloth AI Qwen3.6 MTP variants on a DGX Spark (UD-Q6_K_XL): > 27B → 27B MTP: 8.1 → 18.65 t/s (2.3x faster) > 35B A3B → 35B A3B MTP: 56.91 → 66.52 t/s (+17%) The 27B dense model more than doubled throughput from MTP alone. Free speed is free speed.

2.3x faster. Ran Unsloth AI Qwen3.6 MTP variants on a DGX Spark (UD-Q6_K_XL): > 27B → 27B MTP: 8.1 → 18.65 t/s (2.3x faster) > 35B A3B → 35B A3B MTP: 56.91 → 66.52 t/s (+17%) The 27B dense model more than doubled throughput from MTP alone. Free speed is free speed.

stevibe

24,315 views • 5 months ago

We introduce W.A.L.T, a diffusion model for photorealistic video generation. Our model is a transformer trained on image and video generation in a shared latent space. 🧵👇

We introduce W.A.L.T, a diffusion model for photorealistic video generation. Our model is a transformer trained on image and video generation in a shared latent space. 🧵👇

Agrim Gupta

431,392 views • 2 years ago

Meet VLA-JEPA, the first world model policy in LeRobot 🌍 Most VLAs just learn observation → action. VLA-JEPA also learns action-relevant dynamics: during training, a JEPA world model has to predict upcoming frames in latent space from the model's own actions. Then, at inference, the world model disappears entirely, leaving a standard, fast VLA: Qwen backbone, action head, nothing extra. In our demo it was fine-tuned on just 13 examples and ran in real time on an NVIDIA DGX Spark. On a consumer RTX 3080, it hits 10Hz using under 6GB of VRAM. World-model supervision during training, zero extra cost at inference. That's the whole trick.

Meet VLA-JEPA, the first world model policy in LeRobot 🌍 Most VLAs just learn observation → action. VLA-JEPA also learns action-relevant dynamics: during training, a JEPA world model has to predict upcoming frames in latent space from the model's own actions. Then, at inference, the world model disappears entirely, leaving a standard, fast VLA: Qwen backbone, action head, nothing extra. In our demo it was fine-tuned on just 13 examples and ran in real time on an NVIDIA DGX Spark. On a consumer RTX 3080, it hits 10Hz using under 6GB of VRAM. World-model supervision during training, zero extra cost at inference. That's the whole trick.

LeRobot

42,389 views • 3 months ago

Another major milestone for NATIX 🚀 Researchers at the University of Freiburg trained Orbis 2, a state-of-the-art driving World Model, using NATIX as the largest single data source. Here's why that's a big deal 👇

Another major milestone for NATIX 🚀 Researchers at the University of Freiburg trained Orbis 2, a state-of-the-art driving World Model, using NATIX as the largest single data source. Here's why that's a big deal 👇

NATIX Network

39,507 views • 3 months ago