Daniel Han retweeted
Your agents have spoken. They want more compute. We’ve answered: a new Mac Mini M6 is being raffled off TONIGHT at the Open Source AI Awards. Keep your agents happy. 🏆
I’ll be speaking about why open source AI and open models are essential to keeping innovation accessible, AI eedom to build in everyone’s hands.
Joining me on stage:
• Peter Walker @PeterJ_Walker — @OpenRouter
• Ian Ballantyne @IanBallantyne — @GoogleDeepMind
• Daniel Han @danielhanchen — @UnslothAI
• Irene Solaiman @IreneSolaiman — @huggingface
• Ahmad M. Osman @TheAhmadOsman — @OsmanticAI
• Alex Newman @thedotmack — @Claude_Memory
• Lucas Atkins @latkins — @arcee_ai
• Basil Chatha @realbasilchatha — @latentspacepod, moderating the fireside with Peter
Then we celebrate the builders behind the breakthroughs with @open_intel_fdn.
🏆 Project awards:
The Bedrock Award • The Reward Signal Award • The Guardrails Award • The Reality Check Award • The Mission Control Award • The Glass Box Award
🏆 Individual awards:
The Community Lifeline Award • The Unsung Maintainer Award • The Open Knowledge Award • The Naming Things Is Hard Award
And yes, there are prizes for the “Best OSAI Week Post” on X. Share your highlights using BOTH #OpenSourceAIWeek and #OpenIntelligence to enter.
🍽️ Doors open at 5 PM with hot food
🎤 Program starts at 6 PM sharp
Interest is high and entry is first come, first served. Arrive early!
Registration is required: luma.com/7ebvz0oq
Let’s give the people building open AI their moment.
#OpenSourceAIWeek #OpenIntelligence
I’ll be speaking about why open source AI and open models are essential to keeping innovation accessible, AI eedom to build in everyone’s hands.
Joining me on stage:
• Peter Walker @PeterJ_Walker — @OpenRouter
• Ian Ballantyne @IanBallantyne — @GoogleDeepMind
• Daniel Han @danielhanchen — @UnslothAI
• Irene Solaiman @IreneSolaiman — @huggingface
• Ahmad M. Osman @TheAhmadOsman — @OsmanticAI
• Alex Newman @thedotmack — @Claude_Memory
• Lucas Atkins @latkins — @arcee_ai
• Basil Chatha @realbasilchatha — @latentspacepod, moderating the fireside with Peter
Then we celebrate the builders behind the breakthroughs with @open_intel_fdn.
🏆 Project awards:
The Bedrock Award • The Reward Signal Award • The Guardrails Award • The Reality Check Award • The Mission Control Award • The Glass Box Award
🏆 Individual awards:
The Community Lifeline Award • The Unsung Maintainer Award • The Open Knowledge Award • The Naming Things Is Hard Award
And yes, there are prizes for the “Best OSAI Week Post” on X. Share your highlights using BOTH #OpenSourceAIWeek and #OpenIntelligence to enter.
🍽️ Doors open at 5 PM with hot food
🎤 Program starts at 6 PM sharp
Interest is high and entry is first come, first served. Arrive early!
Registration is required: luma.com/7ebvz0oq
Let’s give the people building open AI their moment.
#OpenSourceAIWeek #OpenIntelligence
3
3
19
3,716
Daniel Han retweeted
NVIDIA releases Nemotron-3.5-Bolt-E4B, a new model that runs locally on 6GB RAM. 💚
Bolt-E4B can run on edge devices like phones and delivers top-tier performance for its size. Trained on Gemma 4, it supports vision.
GGUF: huggingface.co/unsloth/NVIDIA…
Guide: unsloth.ai/docs/models/ne…
Bolt-E4B can run on edge devices like phones and delivers top-tier performance for its size. Trained on Gemma 4, it supports vision.
GGUF: huggingface.co/unsloth/NVIDIA…
Guide: unsloth.ai/docs/models/ne…
Available now: NVIDIA Nemotron 3.5 Bolt E4B 4GB ⚡️
Our newest small open model is built for local assistants, interactive characters and robots that can carry on a conversation and use tools to get things done.
Watch Reachy Mini chat and move, with Bolt handling conversation ing vision. Both models run locally on Jetson Orin Nano 8GB.
Learn more: research.nvidia.com/labs/nemotron/…
88
202
2,282
118,987
Daniel Han retweeted
We plugged Beam into @opencode, asked it to look at @UnslothAI, and create a fine-tuning notebook for the latest and smallest Gemma-4 model on a Text2SQL task – an out of distribution domain for Beam.
The model successfully researches the model cards, Unsloth documentation, and implements the fine-tuning logic and surrounding machinery. As a result, Beam is able to successfully finetune Gemma, increasing Gemma’s accuracy on its heldout test set by 66.5%.
The model successfully researches the model cards, Unsloth documentation, and implements the fine-tuning logic and surrounding machinery. As a result, Beam is able to successfully finetune Gemma, increasing Gemma’s accuracy on its heldout test set by 66.5%.
14
19
191
11,801
We made Flint Unsloth GGUFs - Recursive's new model - it's an MoE that's the best perf for its size!
It works great for tool calls and agentic coding in Unsloth Desktop! We also now have a beta browser + browser-use, skills, more themes and more! Flint can use them all!
It works great for tool calls and agentic coding in Unsloth Desktop! We also now have a beta browser + browser-use, skills, more themes and more! Flint can use them all!
Run Flint locally on 24GB RAM via Unsloth Dynamic GGUFs!
Flint is the strongest MoE model under 50B parameters.
Recursive’s Apache-2.0 model has 39.8B parameters (2.5B active). Run at 125 t/s on Mac, 150 on RTX Spark
GGUF: huggingface.co/unsloth/flint-…
Guide: unsloth.ai/docs/models/fl…
8
10
87
4,720
We made some Unsloth UD3 dynamic quants for Ornith's new 122B release! UD-Q2_K_XL is 44.5GB and retains around 84% top-1% for wikitext general and 89% ish for chat whilst being 82% smaller (BF16 = 249GB)
🥳 Ornith-1.5-122B Unsloth GGUFs are now available!
🧡 Thanks for the awesome work by @UnslothAI
🔗 huggingface.co/unsloth/Ornith…
9
12
55
4,069
We worked on a llama.cpp impl for DS4.1-Flash + made Unsloth dynamic GGUFs for it!
The engram / PLE can be swapped out from MXFP8 to Q6_K to Q5_K and IQ4_NL precision - we found Q6_K approx = MXFP8 and IQ4_NL loses a tiny bit of accuracy.
The engram can stream fully from disk!
The engram / PLE can be swapped out from MXFP8 to Q6_K to Q5_K and IQ4_NL precision - we found Q6_K approx = MXFP8 and IQ4_NL loses a tiny bit of accuracy.
The engram can stream fully from disk!
DeepSeek V4.1-Flash-1017 can now run locally! 🐳
We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%).
Run on a 128GB+ RAM device.
V4.1 Flash 1017 is the strongest open model to date
Guide: unsloth.ai/docs/models/de…
GGUF: huggingface.co/unsloth/DeepSe…
12
19
145
9,720
Unsloth now works with ARM64 + Windows + CUDA on RTX Spark! We made [running, training] X [diffusion, multimodal, language, decision] models all run natively + we optimized it to make it 2.7x faster!
Get it via winget or unsloth.ai/download/windo…
Get it via winget or unsloth.ai/download/windo…
You can now run and train 120B+ parameter models locally on a laptop! NVIDIA RTX Spark is here. 💚
With Unsloth and RTX Spark, train diffusion, decision, multimodal models ~2.7× faster with optimized kernels and binaries for CUDA on ARM. All local with up to 128GB unified memory
3
4
32
2,316
Daniel Han retweeted
how do you fine-tune a decision model on your own data?
in these two notebooks, you'll learn:
• how to prepare the training data
• how to fine-tune our d1-3B decision model
• using @UnslothAI and @huggingface TRL
all within 30 minutes on a free-tier GPU.
unsloth: github.com/Liquid4All/coo…
TRL: github.com/Liquid4All/coo…
ps: how adorable are these halloween kitties? 👻
in these two notebooks, you'll learn:
• how to prepare the training data
• how to fine-tune our d1-3B decision model
• using @UnslothAI and @huggingface TRL
all within 30 minutes on a free-tier GPU.
unsloth: github.com/Liquid4All/coo…
TRL: github.com/Liquid4All/coo…
ps: how adorable are these halloween kitties? 👻
42
241
1,989
84,456
Muse Spark Open from Meta is here! We made Unsloth UD-3 quants for the model with UD-IQ1_S being 176GB
We also did some KLD benchmarks and recovered as much accuracy as we could for the smaller quants!
Muse Spark Open is fully multimodal and works in Unsloth Desktop!
We also did some KLD benchmarks and recovered as much accuracy as we could for the smaller quants!
Muse Spark Open is fully multimodal and works in Unsloth Desktop!
Meta releases Muse Spark Open, a new SOTA open model.
Run on a Mac Studio via Dynamic 2-bit Unsloth GGUFs.
Muse-Spark-Open-816B-A42B supports text, images, video and audio input with 1M context.
GGUF: huggingface.co/unsloth/Muse-S…
Guide: unsloth.ai/docs/models/mu…
5
1
35
2,283
We created some Unsloth UD-3 GGUF quants for NVIDIA's Nemotron 3.5 Super and Bolt model!
Bolt runs great in RTX Spark!
We also made NVFP4 GGUFs for Super!
We also ran KLD benchmarks for both - docs at unsloth.ai/docs/models/ne… and unsloth.ai/docs/models/ne…
Bolt runs great in RTX Spark!
We also made NVFP4 GGUFs for Super!
We also ran KLD benchmarks for both - docs at unsloth.ai/docs/models/ne… and unsloth.ai/docs/models/ne…
NVIDIA releases 2 new models: Nemotron-3.5-Bolt and Nemotron-3.5-Super-VL.
Nemotron-3.5-Bolt-74B-A9B runs on 24GB RAM. It delivers top-tier performance for its size.
Super-VL-120B-A12B runs on 82GB RAM.
Bolt GGUF: huggingface.co/unsloth/NVIDIA…
Guide: unsloth.ai/docs/models/ne…
Super sxzkLo
7
5
41
2,555
Daniel Han retweeted
Yesterday I trained my own Jev style decision model by finetuning Qwen3.5-4B on the LocalLLaMA/typed-decisions dataset with @UnslothAI
added a decision head + trained with 4-bit QLoRA on a single A100 within 15 mins, 4 epochs -> 80% accuracy (2 Epochs are enough though)
Merged theQ4_K_M.
Today its so satisfying to watch it play Flappy Bird locally on a single RTX 4060 via llama.cpp.
250ms latency (faster than most APIs I've tested) in just 4GB VRAM during inference.
What should I finetune + test next?
added a decision head + trained with 4-bit QLoRA on a single A100 within 15 mins, 4 epochs -> 80% accuracy (2 Epochs are enough though)
Merged theQ4_K_M.
Today its so satisfying to watch it play Flappy Bird locally on a single RTX 4060 via llama.cpp.
250ms latency (faster than most APIs I've tested) in just 4GB VRAM during inference.
What should I finetune + test next?
Finetuned Qwen3.5-4B into a decision model like Jev using @UnslothAI.
- 4-bit QLoRA, bf16, single A100, 15 min.
- Train/val loss flatten after 2 epochs so most of the learning happens fast.
- Trained just 64.5M params (2.1%)
- accuracy jumps 30% -> 80% after 4 epochs on Loceval dataset
What should i try next?
14
18
118
11,644
Daniel Han retweeted
Fine-tuning my own decision model on my old gaming rig: @googlegemma Gemma 4 E2B + LoRA in @UnslothAI Studio.
16 GB of VRAM, about 12 minutes for the run in total.
If you've never fine-tuned a model, this is your sign. The barrier is a lot lower than you think.
16 GB of VRAM, about 12 minutes for the run in total.
If you've never fine-tuned a model, this is your sign. The barrier is a lot lower than you think.
19
26
370
35,211
Daniel Han retweeted
Video: Decision model trained in ~2 mins on a single DGX Spark using Unsloth! Qwen3.5-2B as base model.
Just follow instruction here: unsloth.ai/docs/basics/tr…
Just follow instruction here: unsloth.ai/docs/basics/tr…
12
34
326
24,195
We're doing a meetup at Dolores Park, Saturday 2PM - we'll be around the palm trees at the center of the park!
Co-hosted with @arankomatsuzaki @iScienceLuvr @gordic_aleksa
All are welcome!
Co-hosted with @arankomatsuzaki @iScienceLuvr @gordic_aleksa
All are welcome!
We're hosting a meetup at Dolores Park this Saturday from 2 pm!
Co-hosted by @iScienceLuvr (@SophontAI), @gordic_aleksa, @danielhanchen + Mike Chen (@UnslothAI)
See you soon :)
3
7
38
6,117
Daniel Han retweeted
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 del 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.
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 del 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.
40
98
716
59,166
Daniel Han retweeted
This is a goldmine, bookmark worthy.
You can now train your own Decision model with our free notebook! 💡
Qwen3.5-4B will generate decisions instead of text on just 8GB VRAM locally.
Learn to data prep (state, questions, gold answers), train, serve.
Notebook: colab.research.google.com/github/unsloth…
Guide: unsloth.ai/docs/basics/tr…
8
229
2,968
177,470
Daniel Han retweeted
You can now train your own Decision model with our free notebook! 💡
Qwen3.5-4B will generate decisions instead of text on just 8GB VRAM locally.
Learn to data prep (state, questions, gold answers), train, serve.
Notebook: colab.research.google.com/github/unsloth…
Guide: unsloth.ai/docs/basics/tr…
Qwen3.5-4B will generate decisions instead of text on just 8GB VRAM locally.
Learn to data prep (state, questions, gold answers), train, serve.
Notebook: colab.research.google.com/github/unsloth…
Guide: unsloth.ai/docs/basics/tr…
You can now train your own Decision model like Jev locally!
We increased Qwen3.5 0.8B’s aggregate accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM.
Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source Unsloth repo.
We head using Unsloth and LoRA (r=64) for one epoch, increasing downstream accuracy from 30–37% to 78%.
GitHub: github.com/unslothai/unsl…
Guide and Notebooks: unsloth.ai/docs/basics/tr…
111
450
3,757
351,433
Daniel Han retweeted
Develop locally. Create anywhere.
Fine-tune large models, code with AI agents, and deploy your projects on a PC that goes where you go, powered by NVIDIA RTX Spark.
@UnslothAI @LMStudio @code @NousResearch @OpenClaw @perplexity_ai
Fine-tune large models, code with AI agents, and deploy your projects on a PC that goes where you go, powered by NVIDIA RTX Spark.
@UnslothAI @LMStudio @code @NousResearch @OpenClaw @perplexity_ai
6
21
252
203,899
Daniel Han retweeted
More agents. More partners. A more secure foundation. 🤝
With MXC now GA on Windows 11, a growing ecosystem of partners is bringing support for policy-driven agent containment.
See who's building with MXC: msft.it/6019aUbSL
With MXC now GA on Windows 11, a growing ecosystem of partners is bringing support for policy-driven agent containment.
See who's building with MXC: msft.it/6019aUbSL
5
13
108
11,379
We added OS level sandoxing in Unsloth with bwrap (Linux), seatbelt (Mac) and Windows MXC in Unsloth!
Latency per tool call for all is under 100ms. Our software style sandboxing with regex ast checks is 3ms latency as well.
Thanks to Windows for collabing with us on MXC!
Latency per tool call for all is under 100ms. Our software style sandboxing with regex ast checks is 3ms latency as well.
Thanks to Windows for collabing with us on MXC!
Windows now has sandboxing!
Microsoft released an open-source repo, mxc, for sandboxed code execution.
We collaborated with Windows to add mxc OS level sandboxing to Unsloth which adds just <100 ms of overhead.
GitHub: github.com/unslothai/unsl…
Guide: unsloth.ai/docs/new/studi…
6
10
63
3,249