--- title: Unsloth Desktop: what it does and what's new | daily.dev description: Unsloth, known for GGUF dynamic quants, has released a free open-source desktop app in beta for Mac, Windows, and Linux that combines chat, model management,... canonical: https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Unsloth Desktop: what it does and what's new | daily.dev og:description: Unsloth, known for GGUF dynamic quants, has released a free open-source desktop app in beta for Mac, Windows, and Linux that combines chat, model management,... og:url: https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x og:image: https://api.daily.dev/og/posts/fZNfnim8x.png og:image:alt: Unsloth Desktop: what it does and what's new og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Unsloth Desktop: what it does and what's new **[Collections](https://daily.dev/sources/collections)** · 3 min read · 3 upvotes · 0 comments ## Summary Unsloth, known for GGUF dynamic quants, has released a free open-source desktop app in beta for Mac, Windows, and Linux that combines chat, model management, fine-tuning (QLoRA, LoRA, full fine-tuning, continued pretraining), agent connections, and model import from other local runners like Ollama and LM Studio. A recent update adds multi-user accounts, revamped Docker integration, a custom Jupyter Notebook, RDNA1/RDNA2 AMD GPU support, ARM64 Windows CUDA support, faster GRPO training, and FP8/INT8 image diffusion support with roughly 2x faster inference. ## Content Unsloth has had a busy stretch. The team launched a free, open-source desktop app in beta for Mac, Windows, and Linux. Then Pillar Security found that just selecting a model in Unsloth Studio could execute arbitrary code on your machine. ## What the desktop app does Unsloth is best known for its GGUF dynamic quants. The new app puts several jobs into one workspace: chat, model management, fine-tuning (QLoRA, LoRA, full fine-tuning, continued pretraining), agent connections, and remote access. A few things stand out. You get deep per-model configuration options. Unsloth's own quants are available from day zero. You can run parallel chat sessions. And it imports models you already have from other local runners, so you don't have to re-import anything. One reviewer who tried it said he no longer wants to go back to LM Studio, Ollama, or AnythingLLM. @danielhanchen has also listed recent additions: multi-user accounts in Unsloth Desktop, a revamped Docker setup, a custom Jupyter Notebook with themes, titles, and expandable cells, RDNA1 and RDNA2 support, ARM64 Windows CUDA support, faster GRPO, and FP8/INT8 image diffusion support for 2x faster inference. ## The code-execution problem Pillar Security found that in Unsloth Studio, picking a model for inspection was enough to trigger Python code execution. No inference and no weight loading were needed. The cause was a default. During a routine metadata check on a model's config.json, Unsloth set Hugging Face's trust_remote_code option to true. That option tells the library to download and run Python code shipped in the model's repository. So a malicious repo could run code on your machine the moment you clicked it in the picker. The potential exposure is broad: Hugging Face tokens, SSH keys, cloud credentials, and proprietary training data. ## The disclosure dispute Unsloth's maintainers declined to publish a security advisory or request a CVE, pointing to Studio's beta status. Pillar disagrees with that framing. The vulnerable code ships in the generally available PyPI package, which anyone can install with a plain pip install, so the beta label on the UI doesn't limit who is exposed. ## The fix Unsloth fixed it in version 2026.6.9. The release disables arbitrary model loading from Hugging Face and stops trusting remote code from local model files. Pillar independently confirmed that the fix closes both attack paths, the Hugging Face one and the local-directory one. ## What to do Upgrade, even if you never launch Studio, since the vulnerable code is in the base package. Then check your own LLM workflows for places where trust_remote_code=True is set without a real need for it. Convenience defaults like this one are easy to forget until they bite. ## Questions this post answers ### What new features did Unsloth Desktop just add? Unsloth Desktop's latest update adds multi-user accounts for shared or server use, a revamped Docker integration, a custom Jupyter Notebook with custom themes and expandable cells, RDNA1 and RDNA2 AMD GPU support, ARM64 Windows CUDA support, faster GRPO training, and FP8/INT8 image diffusion support giving roughly 2x faster inference. _Explore ongoing local-LLM tooling updates like this one on daily.dev._ ### Can Unsloth Desktop import models already downloaded through Ollama or LM Studio? Yes, Unsloth Desktop can import existing models from other local runners like Ollama and LM Studio without needing to redownload them, so users switching tools keep their existing model library intact. _Developers weighing local LLM tool migrations can track compatibility details like this on daily.dev._ ### Does Unsloth Desktop support fine-tuning methods like QLoRA and full fine-tuning? Yes, Unsloth Desktop supports QLoRA, LoRA, full fine-tuning, and continued pretraining directly within the app, alongside chat, model management, and agent connections, consolidating tools that previously required separate apps like LM Studio, Ollama, and AnythingLLM. _Anyone comparing local fine-tuning setups can follow tool consolidation stories like this on daily.dev._ ## Similar posts on daily.dev - [Malicious Hugging Face model masquerading as OpenAI release hits 244K downloads](https://daily.dev/posts/malicious-hugging-face-model-masquerading-as-openai-release-hits-244k-downloads-xftpmcpz1) · CSO Online · 0 upvotes · 0 comments - [Remote Code Execution With Modern AI/ML Formats and Libraries](https://daily.dev/posts/remote-code-execution-with-modern-ai-ml-formats-and-libraries-kzzcz2ymw) · Unit 42 · 0 upvotes · 0 comments - [Hugging Face and ClawHub compromised with hundreds of malicious AI models and agent skills as supply chain attacks target AI infrastructure](https://daily.dev/posts/hugging-face-and-clawhub-compromised-with-hundreds-of-malicious-ai-models-and-agent-skills-as-supply-xjnujrk1y) · The Next Web · 4 upvotes · 0 comments - [Hugging Face Hack Lessons for Cyber Defenders](https://daily.dev/posts/hugging-face-hack-lessons-for-cyber-defenders-z3vltbx1w) · Dark Reading · 0 upvotes · 0 comments --- Tags: [#deep-learning](https://daily.dev/tags/deep-learning), [#ollama](https://daily.dev/tags/ollama), [#lm-studio](https://daily.dev/tags/lm-studio) [View this post on daily.dev](https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Unsloth Desktop: what it does and what's new","url":"https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x"},"datePublished":"2026-09-23T23:04:34.742Z","dateModified":"2026-09-30T15:24:49.749Z","description":"Unsloth, known for GGUF dynamic quants, has released a free open-source desktop app in beta for Mac, Windows, and Linux that combines chat, model management,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9382b84cae4bf37bf993606b084eeb63?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/9382b84cae4bf37bf993606b084eeb63?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"deep-learning,ollama,lm-studio","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Unsloth Desktop: what it does and what's new"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/unsloth-desktop-what-it-does-and-what-s-new-fznfnim8x#faq","mainEntity":[{"@type":"Question","name":"What new features did Unsloth Desktop just add?","acceptedAnswer":{"@type":"Answer","text":"Unsloth Desktop's latest update adds multi-user accounts for shared or server use, a revamped Docker integration, a custom Jupyter Notebook with custom themes and expandable cells, RDNA1 and RDNA2 AMD GPU support, ARM64 Windows CUDA support, faster GRPO training, and FP8/INT8 image diffusion support giving roughly 2x faster inference. Explore ongoing local-LLM tooling updates like this one on daily.dev."}},{"@type":"Question","name":"Can Unsloth Desktop import models already downloaded through Ollama or LM Studio?","acceptedAnswer":{"@type":"Answer","text":"Yes, Unsloth Desktop can import existing models from other local runners like Ollama and LM Studio without needing to redownload them, so users switching tools keep their existing model library intact. Developers weighing local LLM tool migrations can track compatibility details like this on daily.dev."}},{"@type":"Question","name":"Does Unsloth Desktop support fine-tuning methods like QLoRA and full fine-tuning?","acceptedAnswer":{"@type":"Answer","text":"Yes, Unsloth Desktop supports QLoRA, LoRA, full fine-tuning, and continued pretraining directly within the app, alongside chat, model management, and agent connections, consolidating tools that previously required separate apps like LM Studio, Ollama, and AnythingLLM. Anyone comparing local fine-tuning setups can follow tool consolidation stories like this on daily.dev."}}]} ```