{"id":14893,"date":"2026-11-17T11:03:16","date_gmt":"2026-11-17T02:03:16","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=14893"},"modified":"2026-11-17T11:12:46","modified_gmt":"2026-11-17T02:12:46","slug":"unsloth-v0-1-914-beta-release-2","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-914-beta-release-2\/","title":{"rendered":"Unsloth v0.1.914-beta Released with Glass Theme and Speedups"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/github.com\/unslothai\/unsloth\">unslothai\/unsloth<\/a><\/td>\n<\/tr>\n<tr>\n<td>Version<\/td>\n<td><a href=\"https:\/\/github.com\/unslothai\/unsloth\/releases\/tag\/v0.1.914-beta\">v0.1.914-beta<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-03<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>Unsloth is a UI tool for running and training LLMs (Large Language Models) and diffusion models locally. In this update (v0.1.914-beta), UI customizability has been significantly improved with the introduction of a new &#8220;Glass theme&#8221;.<\/p>\n<p>In addition, generation speeds for Llama and Qwen3 models in NVIDIA GPU environments have been substantially boosted, making this a highly valuable update for users who prioritize inference performance.<\/p>\n<h2>Key Changes<\/h2>\n<h3>Enhanced UI Customization<\/h3>\n<p>A new &#8220;Glass theme&#8221; has been introduced. Five styles are available, allowing you to adjust transparency and choose whether or not to display a background image. The macOS version of Unsloth Desktop supports a native glass effect that allows your desktop wallpaper to show through the window.<\/p>\n<p>It is also now possible to customize greeting text, icons, and images, as well as configure code fonts, line numbers, word wrap, and ligatures. These can be adjusted while checking them with a live preview.<\/p>\n<h3>Improved Inference Performance<\/h3>\n<p>When using NVIDIA GPUs on Linux or WSL, single-GPU generation speeds for Llama and Qwen3 models have increased in Unsloth&#8217;s Transformers backend. In tests using an A100, 4-bit Qwen3-8B recorded an approximate 3.9x speedup, jumping from the previous 25.7 tokens\/s to 99.1 tokens\/s.<\/p>\n<h3>Chat Feature Improvements<\/h3>\n<p>The new chat navigator allows you to preview past messages, bookmark turns, and jump within long conversations. It also supports pasting content in Markdown format (including links, lists, and spreadsheet tables).<\/p>\n<p>Display support for Right-to-Left (RTL) languages such as Arabic, Persian, Hebrew, and Urdu has been added. Note that code portions maintain their original direction.<\/p>\n<h3>Expanded Training Options<\/h3>\n<p>Even in environments without GPU availability, you can now fine-tune text models directly on the CPU from the Train page. Additionally, for audio-vision models such as Gemma 3n and Gemma 4, it is now possible to train using a combination of audio and images within the same dataset.<\/p>\n<p>In multi-GPU environments, you can use <code>UNSLOTH_PARALLEL<\/code> to configure combinations of data, tensor, pipeline, and context parallelism. Furthermore, using <code>unsloth.tile_mlp(model)<\/code> reduces MLP activation memory during training.<\/p>\n<h3>Image, Video, and Audio Generation Optimization<\/h3>\n<p>SDXL can now render with a VRAM budget of under 4 GiB (tested on T4 and L4 GPUs). For video models, repeated loading has been accelerated by reusing cached INT8 weights. Transcription of long audio files of up to 4 hours is also available using audio.cpp (subject to upload and storage limits).<\/p>\n<h2>Supported Models and Hardware<\/h2>\n<p>This update brings support and improvements for the following models and hardware.<\/p>\n<h3>Supported Models<\/h3>\n<ul>\n<li><strong>Llama \/ Qwen3<\/strong>: Generation speed improvements have been confirmed with the Transformers backend on NVIDIA GPUs (Linux\/WSL environments).<\/li>\n<li><strong>DeepSeek-V4.1<\/strong>: GGUF formats are provided that show strong performance across Terminal Bench, DeepSWE, and AutomationBench items in benchmarks with a 512K context.<\/li>\n<li><strong>SDXL<\/strong>: Rendering with a VRAM budget of under 4 GiB is now possible on T4 and L4 GPUs.<\/li>\n<li><strong>Wan2.2 A14B<\/strong>: System memory usage is reduced when building INT8 weights for offloaded generation.<\/li>\n<li><strong>Gemma 3n \/ Gemma 4<\/strong>: Fine-tuning audio-vision models is possible by combining audio and images in the same dataset.<\/li>\n<\/ul>\n<h3>Hardware and Accelerators<\/h3>\n<ul>\n<li><strong>macOS<\/strong>: Unsloth Desktop supports a native glass effect (allowing the desktop to show through). Experimental &#8220;Fast Verify&#8221; using MLX also enables fast generation on compatible MLX speculative decoding models.<\/li>\n<li><strong>NVIDIA GPU<\/strong>: Generation performance on Linux or WSL environments has been improved.<\/li>\n<li><strong>CPU<\/strong>: Text model training can now be executed directly from the Train page even when GPUs are unavailable.<\/li>\n<\/ul>\n<h2>How to Get It<\/h2>\n<p>To install or update to the latest development version (main branch) of Unsloth Studio (Web UI), run the following steps.<\/p>\n<h3>macOS, Linux, and WSL<\/h3>\n<pre><code class=\"language-bash\">git clone https:\/\/github.com\/unslothai\/unsloth\ncd unsloth\n.\/install.sh --local\nunsloth studio -p 8888\n<\/code><\/pre>\n<p>To update, use the following commands:<\/p>\n<pre><code class=\"language-bash\">cd unsloth &amp;&amp; git pull\n.\/install.sh --local\nunsloth studio -p 8888\n<\/code><\/pre>\n<h3>Windows PowerShell<\/h3>\n<pre><code class=\"language-powershell\">git clone https:\/\/github.com\/unslothai\/unsloth.git\ncd unsloth\nSet-ExecutionPolicy -Scope Process -ExecutionPolicy RemoteSigned\n.\\install.ps1 --local\nunsloth studio -p 8888\n<\/code><\/pre>\n<p>To update, use the following commands:<\/p>\n<pre><code class=\"language-powershell\">cd unsloth; git pull\n.\\install.ps1 --local\nunsloth studio -p 8888\n<\/code><\/pre>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-918-beta-release\/\">Unsloth v0.1.918-beta Adds Computer Use, MCP Game Engine Support<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-917-beta-released\/\">Unsloth v0.1.917-beta Released with Major Performance Boosts<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-916-beta-released\/\">Unsloth v0.1.916-beta Released: Performance &amp; Multi-GPU Updates<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-915-beta-released\/\">Unsloth v0.1.915-beta Released with Browser Use &amp; Qwen3.8<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Follow this tool<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-unsloth-en\/\">Unsloth overview and release history (89 releases tracked)<\/a><\/li>\n<li><strong>Other quantization, model formats and fine-tuning<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-ggml-en\/\">ggml<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-exllamav3-en\/\">ExLlamaV3<\/a><\/li>\n<li><strong>How to read Terminal-Bench<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-benchmarks-en\/\">Benchmark glossary<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/github.com\/unslothai\/unsloth\/releases\/tag\/v0.1.914-beta\">unslothai\/unsloth v0.1.914-beta Release Notes<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Unsloth v0.1.914-beta is out, featuring a new Glass theme, faster Llama and Qwen3 generation on NVIDIA GPUs, and expanded training options.<\/p>\n","protected":false},"author":1,"featured_media":14892,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[105],"tags":[163,4538,375,509,1547,2145],"class_list":["post-14893","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-engines-and-tools","tag-gguf-en","tag-llama-en","tag-qwen3-en","tag-unsloth-en","tag-verified","tag--en"],"lang":"en","translations":{"en":14893,"ja":14891},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14893","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=14893"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14893\/revisions"}],"predecessor-version":[{"id":18032,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14893\/revisions\/18032"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/14892"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=14893"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=14893"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=14893"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}