{"id":14363,"date":"2026-11-17T11:03:09","date_gmt":"2026-11-17T02:03:09","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=14363"},"modified":"2026-11-17T11:12:31","modified_gmt":"2026-11-17T02:12:31","slug":"unsloth-v0-1-913-beta-release","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-913-beta-release\/","title":{"rendered":"Unsloth v0.1.913-beta Released: Faster Updates and Image Gen"},"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\/v0.1.913-beta\">v0.1.913-beta<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-10-30<\/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 local UI tool for running and training LLMs (Large Language Models) and diffusion models. It supports a diverse range of models including GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, and FLUX.<\/p>\n<p>In this update, UI\/UX improvements and customization features have been added, alongside a significant reduction in app update times. Notably, generation performance for Linux apps and Qwen-Image has improved in environments with limited VRAM capacity.<\/p>\n<h2>Breaking Changes and Deprecations<\/h2>\n<p>There is a change in behavior regarding app updates.<\/p>\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 style=\"text-align: left;\">Item<\/th>\n<th style=\"text-align: left;\">Behavior up to previous version<\/th>\n<th style=\"text-align: left;\">New version (v0.1.913-beta)<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"text-align: left;\">App updates<\/td>\n<td style=\"text-align: left;\">Executes Python precompilation<\/td>\n<td style=\"text-align: left;\">Skips Python precompilation<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>Due to this change, the time required to complete app updates is reduced by about half, but the first launch time after updating is longer than before (v0.1.912 and earlier). While this is a major benefit for users who update frequently, users who prioritize launch speed should take note.<\/p>\n<h2>Key Changes<\/h2>\n<h3>UI\/UX Customization and Improvements<\/h3>\n<p>You can now customize the chat screen background. Additionally, under &#8220;Settings &gt; Chat&#8221;, you can now configure &#8220;Composer at the bottom&#8221; to place the chat box at the bottom of the screen during new chats. Furthermore, interface cleanups have been made, such as completely hiding the greeting screen and displaying chat cards more compactly in the library.<\/p>\n<h3>Image and Video Generation Performance Improvements<\/h3>\n<p>Rendering speed for Qwen-Image has improved by up to 2x in environments where VRAM (video memory) is limited to 16 GB. Qwen-Image-Edit-2511 also achieves about a 1.3x speedup. Furthermore, FLUX.2-klein-9B can now be loaded and rendered even in Linux environments equipped with 12.7 GB of system RAM and a compatible GPU.<\/p>\n<h3>Application Responsiveness and Efficiency<\/h3>\n<p>Smoothness when streaming long responses has improved. For example, in parsing an 83 KB response containing invalid math formatting (<code>$$<\/code>), processing that previously took 71 seconds has been reduced to 0.7 seconds. Additionally, peak memory usage when previewing a 1 GB local Parquet file was significantly reduced from 1.5 GB to about 300 MB.<\/p>\n<h3>Optimizations for Mac Users<\/h3>\n<p>On Apple Silicon environments, Claude Code, Codex, and OpenCode requests can now share MLX batches. Tests report that four simultaneous requests completed up to 2.3x faster. Additionally, by enabling the experimental &#8220;Neural Engine Prefill&#8221; feature, prompt processing is accelerated by up to 31% in tests with Qwen3.8-27B-4bit on M3 Max (this is opt-in and may slightly alter responses).<\/p>\n<h3>Training Feature Enhancements<\/h3>\n<p>For MoE (Mixture of Experts) models such as Qwen3.6-35B-A3B, compatible LoRA and QLoRA jobs can now be simultaneously trained through the Fine-tuning API with their respective adapters. Also, single-column text datasets are now automatically recognized, eliminating the need for manual column mapping.<\/p>\n<h2>Supported Models and Hardware<\/h2>\n<p>In this update, support and improvements have been made for the following models and hardware.<\/p>\n<h3>Newly Supported Models<\/h3>\n<p>Unsloth Dynamic v3.1 GGUF (preview version) for DeepSeek-V4.1-Flash-1017 has been added. This model is reported to have the highest accuracy for its size across almost all benchmarks.<\/p>\n<h3>Image and Video Generation Model Improvements<\/h3>\n<ul>\n<li><strong>Qwen-Image<\/strong>: Rendering speed under a 16 GB VRAM limit has improved by up to 2x.<\/li>\n<li><strong>Qwen-Image-Edit-2511<\/strong>: Achieves about a 1.3x speedup under the same conditions.<\/li>\n<li><strong>MiniMax-H3<\/strong>: Clip completion speed in memory-limited NVIDIA environments has improved by about 1.7x.<\/li>\n<li><strong>FLUX.2-klein-9B<\/strong>: Can now be loaded and rendered even in Linux environments equipped with 12.7 GB of system RAM and a compatible GPU.<\/li>\n<\/ul>\n<h3>Hardware-Specific Optimizations<\/h3>\n<ul>\n<li><strong>Linux (NVIDIA, X11, driver 610+)<\/strong>: Scrolling through long chats is smoother. In high-resolution tests, the frame rate has improved from about 17 fps to 60 fps.<\/li>\n<li><strong>Apple Silicon (Mac)<\/strong>: With the experimental &#8220;Neural Engine Prefill&#8221;, prompt processing is accelerated by up to 31% in tests with Qwen3.8-27B-4bit on M3 Max (opt-in setting required).<\/li>\n<\/ul>\n<h2>How to Get It<\/h2>\n<p>For developers or those building from the latest source (Nightly\/Experimental), update using the following steps.<\/p>\n<h3>For macOS, Linux, and WSL<\/h3>\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>For Windows PowerShell<\/h3>\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-v0-1-915-beta-released\/\">Unsloth v0.1.915-beta Released with Browser Use &amp; Qwen3.8<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-0-1-912-beta-updates\/\">Unsloth 0.1.912-beta: Multi-LoRA Batching and Faster Startup<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/unsloth-v0-1-911-beta-released-2\/\">Unsloth v0.1.911-beta Released with Scheduled Tasks and More<\/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<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/github.com\/unslothai\/unsloth\/releases\/tag\/v0.1.913-beta\">unslothai\/unsloth v0.1.913-beta release<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Unsloth v0.1.913-beta is released with faster app updates, improved Qwen-Image generation on 16GB VRAM, and Apple Silicon optimizations.<\/p>\n","protected":false},"author":1,"featured_media":14362,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[105],"tags":[163,111,977,165,509,1547],"class_list":["post-14363","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-engines-and-tools","tag-gguf-en","tag-lora-en","tag-mlx-en","tag-moe-en","tag-unsloth-en","tag-verified"],"lang":"en","translations":{"en":14363,"ja":14361},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14363","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=14363"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14363\/revisions"}],"predecessor-version":[{"id":18012,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14363\/revisions\/18012"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/14362"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=14363"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=14363"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=14363"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}