{"id":11089,"date":"2026-08-14T23:53:42","date_gmt":"2026-08-14T15:53:42","guid":{"rendered":"https:\/\/infernews.com\/blog\/qwen3-8-27b\/"},"modified":"2026-08-15T14:13:00","modified_gmt":"2026-08-15T06:13:00","slug":"qwen3-8-27b","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/qwen3-8-27b\/","title":{"rendered":"Qwen3.8 \u7cfb\u5217\u7684 27B \u958b\u653e\u6b0a\u91cd\u6a21\u578b"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/pasted-d8d1cf832401.jpg\" alt=\"Og image\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u7531 <strong>Qwen3.5<\/strong> \u7684\u67b6\u69cb\u57fa\u790e\u5ef6\u4f38\u800c\u4f86\uff0cQwen3.8-27B\u628a\u9577\u6d41\u7a0b\u63a8\u7406\u3001Agentic \u4efb\u52d9\u540c\u539f\u751f\u8996\u89ba\u7406\u89e3\u653e\u5165\u540c\u4e00\u500b 27B dense \u6a21\u578b\u3002\u4f7f\u7528\u6642\u6700\u76f4\u63a5\u7684\u5dee\u7570\uff0c\u662f\u5b83\u4e0d\u53ea\u770b\u5716\u7b54\u984c\uff0c\u4ea6\u91dd\u5c0d\u7de8\u7a0b\u3001\u5c08\u696d\u5de5\u4f5c\u3001\u7814\u7a76\u8207\u591a\u6b65\u9a5f\u4efb\u52d9\u5b8c\u6210\u5ea6\u4f5c\u5f37\u5316\uff0c\u4e26\u4fdd\u7559\u53ef\u8abf\u6574\u7684 thinking control\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6a21\u578b\u5c64\u9762\u63a1\u7528 <strong>Causal Language Model with Vision Encoder<\/strong>\uff0c\u5c6c\u65bc\u7d93\u904e pre-training \u8207 post-training \u7684 image-text-to-text \u6a21\u578b\u300227B \u53c3\u6578\u300164 \u5c64\u3001hidden size 5120\uff0c\u4ee5\u53ca\u6df7\u5408 <strong>Gated DeltaNet<\/strong> \u8207 <strong>Gated Attention<\/strong> \u7684 hidden layout\uff0c\u53cd\u6620\u5b83\u4e0d\u662f\u55ae\u7d14\u6cbf\u7528\u50b3\u7d71 dense Transformer \u5806\u758a\uff0c\u800c\u662f\u91dd\u5c0d\u9577\u7a0b\u4f9d\u8cf4\u8207\u6548\u7387\u4f5c\u7d50\u69cb\u8abf\u6574\u3002<\/p>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"Qwen 3.8 27B BLOWS MY MIND! Best Local AI Model Yet! Basically Opus Locally! (Fully Tested)\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_J_aqblUWj4k\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJ_aqblUWj4k%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/J_aqblUWj4k\" \/><meta itemprop=\"duration\" content=\"PT19M14S\" \/><meta itemprop=\"uploadDate\" content=\"2026-08-15T05:42:26Z\" \/><\/div><div id=\"lyte_J_aqblUWj4k\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJ_aqblUWj4k%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">Qwen 3.8 27B BLOWS MY MIND! Best Local AI Model Yet! Basically Opus Locally! (Fully Tested)<\/div><\/div><button tabindex=\"0\" class=\"play\"><\/button><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/J_aqblUWj4k\" rel=\"nofollow\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJ_aqblUWj4k%2F0.jpg\" alt=\"Qwen 3.8 27B BLOWS MY MIND! Best Local AI Model Yet! Basically Opus Locally! (Fully Tested)\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"\ud83d\udea8 Everything you\u2019re about to see, I benchmarked using the tool I made. If you want to run these tests yourself or test models on whatever you actually use them for! \u2192 https:\/\/www.woaibench.ai\/ Qwen 3.8 27B might be the BEST local AI model I&#039;ve tested yet! \ud83e\udd2f In this video, I fully test the new open-weight Qwen3.8-27B from Alibaba to see just how close a 27B model running locally can get to frontier models like Claude Opus. \ud83d\udd17 My Links: Sponsor a Video or Do a Demo of Your Product, Contact me: intheworldzofai@gmail.com \ud83d\udd25 Become a Patron (Private Discord): https:\/\/patreon.com\/WorldofAi \ud83e\udde0 Follow me on Twitter: https:\/\/twitter.com\/intheworldofai \ud83d\udea8 Subscribe To The SECOND Channel: https:\/\/www.youtube.com\/@UCYwLV1gDwzGbg7jXQ52bVnQ \ud83d\udc69\ud83c\udffb\u200d\ud83c\udfeb Learn to code with Scrimba \u2013 from fullstack to AI https:\/\/scrimba.com\/?via=worldofai (20% OFF) \ud83d\udea8 Subscribe To The FREE AI Newsletter For Regular AI Updates: https:\/\/intheworldofai.com\/ \ud83d\udc7e Join the World of AI Discord! : https:\/\/discord.gg\/NPf8FCn4cD Something coming soon :) https:\/\/www.skool.com\/worldofai-automation [Must Watch]: OpenAI&#039;s GPT-6 Astra WILL BE AGI! Greatest AI Model Ever!: https:\/\/youtu.be\/KbYio-N8_LU Qwen 3.8 Max IS OUT! Best Open Model? (Fully Tested): https:\/\/youtu.be\/vmLwsoVRo30 DeepSeek V4 Flash GA IS INCREDIBLE! Powerful, Cheap, &amp; Fast! (Fully Tested): https:\/\/youtu.be\/wT42SgaOPK4 \ud83d\udccc LINKS &amp; RESOURCES World of AI Benchmark: https:\/\/www.woaibench.ai\/ AI NEWS: https:\/\/x.com\/intheworldofai Blog: https:\/\/x.com\/Alibaba_Qwen\/status\/2088280182356611304 Weights: https:\/\/huggingface.co\/collections\/Qwen\/qwen38 Model Scope: https:\/\/www.modelscope.cn\/collections\/Qwen\/Qwen38 https:\/\/x.com\/loktar00\/status\/2088305773109465458 https:\/\/x.com\/loktar00\/status\/2088329789434221034 https:\/\/x.com\/MaziyarPanahi\/status\/2088304775444013280?s=20 https:\/\/x.com\/quimedesu\/status\/2088373049502351731?s=20 https:\/\/x.com\/WescheNex1q\/status\/2088309142565831094 I&#039;m running Qwen3.8-27B completely locally on my RTX 4090 with 24GB VRAM, using an Unsloth Dynamic 4-bit quant through Open WebUI. And the results genuinely blew my mind. \ud83d\udd25 In this video: \u2022 Qwen3.8-27B real-world benchmarks \u2022 Agentic &amp; long-horizon coding tests \u2022 One-shot GTA 6-style game generation \u2022 Three.js &amp; frontend development tests \u2022 Vision &amp; multimodal capabilities \u2022 Local inference performance \u2022 Unsloth Dynamic GGUF quantization \u2022 4-bit versions around ~17GB \u2022 How you can run Qwen3.8 locally \u2022 Qwen3.8 vs Claude Opus-level coding \u2022 Whether this is the BEST local AI model yet The craziest part is just how much capability Alibaba has packed into 27 billion parameters. You can run this model on consumer hardware, keep everything local, avoid API costs, and still get seriously impressive coding and agentic performance. If you&#039;re into local AI, open-source\/open-weight models, AI coding agents, or running LLMs on your own hardware, Qwen3.8-27B is absolutely one to watch. Subscribe for more AI model tests, benchmarks, local AI tutorials, and the latest AI news! \ud83d\ude80 [Time Stamps]: 0:00 - Intro 1:51 - Best Local Model 2:00 - Benchmarks 3:02 - Beats Opus 4.6 3:45 - Speed 4:10 - Local Guide 5:46 - My Setup 7:03 - Call of Duty Zombies Clone 9:04 - Design Taste 9:43 - NVIDIA Landing Page 10:30 - Frontend Demo 10:59 - MacOS Clone 12:13 - Multimodal Demo 13:26 - Web Dev 13:44 - SVG Demos 15:16 - 3D Product Viewer 15:41 - Worst Gen? 16:12 - Three.js Demo 17:23 - GTA Clone 17:59 - Two Cents #Qwen #Qwen38 #Qwen3827B #LocalAI #OpenSourceAI #AI #LLM #Alibaba #AICoding #CodingAI #Claude #ClaudeOpus #OpenWebUI #Unsloth #GGUF #RTX4090 #NVIDIA #AgenticAI #AIAgents #LocalLLM #MachineLearning #GenerativeAI #ArtificialIntelligence #VibeCoding Tags: Qwen 3.8 27B, Qwen3.8, Qwen3.8 27B, Qwen 3.8, Qwen 27B, Alibaba Qwen, Qwen AI, Qwen3, local AI, best local AI model, best local LLM, local LLM, open source AI, open weight AI, Claude Opus locally, Qwen vs Claude, Qwen vs Opus, Claude Opus, AI coding, coding AI, agentic coding, AI coding agent, AI agents, agentic AI, Unsloth, Unsloth Qwen, Unsloth Dynamic, GGUF, Qwen GGUF, Qwen 4 bit, 4 bit quantization, Open WebUI, Ollama, LM Studio, RTX 4090 AI, 24GB VRAM LLM, run AI locally, run Qwen locally, Qwen local setup, local AI 2026, best AI model 2026, open source LLM, multimodal AI, vision language model, VLM, GTA 6 AI, vibe coding, Three.js AI, World of AI\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<p class=\"wp-block-paragraph\">\u5c0d\u958b\u767c\u8005\u800c\u8a00\uff0c\u8f03\u6709\u7528\u7684\u662f\u63a8\u7406\u884c\u70ba\u63a7\u5236\uff1athinking mode \u9810\u8a2d\u958b\u555f\uff0c\u53ef\u6309\u8acb\u6c42\u95dc\u9589\uff1b<code>reasoning_effort<\/code> \u53ef\u8abf\u6574\u63a8\u7406\u6df1\u5ea6\uff1b<code>preserve_thinking<\/code> \u53ef\u4fdd\u7559\u6b77\u53f2\u8a0a\u606f\u4e2d\u7684 reasoning context\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u57fa\u790e\u6a21\u578b\u53ef\u78ba\u8a8d\u70ba <strong>Qwen3.5<\/strong> \u67b6\u69cb\u7cfb\u7d71\u4e0a\u7684\u5f8c\u8a13\u7df4\u7248\u672c\uff0c\u800c\u975e\u7368\u7acb\u53e6\u8d77\u7210\u7076<\/li>\n\n\n\n<li>\u539f\u751f\u652f\u63f4\u5716\u7247\u8207\u5f71\u7247\u7406\u89e3\uff0c\u5b9a\u4f4d\u6bd4\u7d14\u6587\u5b57\u6a21\u578b\u66f4\u8cbc\u8fd1 Agentic \u8207\u591a\u6a21\u614b\u5de5\u4f5c\u6d41<\/li>\n\n\n\n<li>\u76f8\u5bb9 <strong>Transformers\u3001vLLM\u3001SGLang\u3001TokenSpeed<\/strong>\uff0c\u65b9\u4fbf\u63a5\u5165\u73fe\u6709\u5806\u758a<\/li>\n\n\n\n<li>\u9801\u9762\u672a\u63d0\u4f9b <strong>GGUF \u683c\u5f0f<\/strong>\u3001\u91cf\u5316\u6a94\u540d\u3001mmproj\u3001Ollama\uff0fllama.cpp\uff0fLM Studio \u5efa\u8b70\u914d\u7f6e<\/li>\n\n\n\n<li>\u4ea6\u672a\u898b <strong>MTP draft speculation<\/strong>\u3001v2 \u6a94\u540d\u8b8a\u66f4\u6216 chat template \u6ce8\u610f\u4e8b\u9805\u7684\u5177\u9ad4\u8aaa\u660e<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u8207\u540c\u7cfb\u524d\u4ee3\u76f8\u6bd4\uff0cQwen3.8\u4e3b\u6253\u7684\u662f\u53ef\u9760\u5b8c\u6210\u6574\u6bb5\u4efb\u52d9\uff0c\u800c\u4e0d\u53ea\u662f\u55ae\u8f2a\u56de\u7b54\u66f4\u8070\u660e\uff1b\u8207\u4e00\u822c\u8996\u89ba\u8a9e\u8a00\u6a21\u578b\u76f8\u6bd4\uff0c\u5b83\u66f4\u5f37\u8abf environment feedback \u4e0b\u7684\u81ea\u4e3b\u898f\u5283\u3002\u76ee\u524d\u516c\u958b\u7684\u662f Transformers \u683c\u5f0f\u6b0a\u91cd\u8207\u8a2d\u5b9a\u6a94\uff0c\u672a\u8db3\u4ee5\u76f4\u63a5\u5224\u65b7\u672c\u5730\u91cf\u5316\u90e8\u7f72\u9580\u6abb\uff0c\u6240\u4ee5\u786c\u4ef6\u9700\u6c42\u3001\u63a8\u85a6\u91cf\u5316\u7b49\u7d1a\u8207\u4e0d\u540c\u91cf\u5316\u4e4b\u9593\u7684\u53d6\u6368\uff0c\u73fe\u968e\u6bb5\u53ea\u80fd\u7b49\u5f8c\u7e8c\u6a21\u578b\u6a94\u6216\u793e\u7fa4\u79fb\u690d\u7248\u672c\u88dc\u4e0a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/huggingface.co\/Qwen\/Qwen3.8-27B\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Qwen3.8-27B\u4ee5 27B \u5bc6\u96c6\u6a21\u578b\u540c\u6642\u8655\u7406\u6587\u5b57\u3001\u5716\u7247\u8207\u5f71\u7247\uff0c\u91cd\u9ede\u4e0d\u53ea\u5728\u8b58\u5225\u5167\u5bb9\uff0c\u800c\u662f\u66f4\u7a69\u5b9a\u5b8c\u6210\u591a\u6b65\u9a5f\u4efb\u52d9\u3002\u5b83\u5ef6\u7e8c Qwen3.5 \u67b6\u69cb\uff0c\u4f46\u9801\u9762\u516c\u958b\u7684\u672c\u5730\u91cf\u5316\u8207 GGUF \u8cc7\u8a0a\u4ecd\u672a\u9f4a\u3002<\/p>\n","protected":false},"author":8,"featured_media":11088,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[116,76,149,119,127,185,137,146],"tags":[],"class_list":["post-11089","post","type-post","status-publish","format-standard","hentry","category-agentic","category-76","category-149","category-119","category-127","category-qwen","category-api","category-146"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11089","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=11089"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11089\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/11088"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=11089"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=11089"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=11089"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}