--- title: QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But... description: A hands-on benchmark comparing Qwen 3.6 35B (A3B) with and without Multi-Token Prediction (MTP) enabled, running locally on an Apple M5 Max via LM Studio.... canonical: https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It? | daily.dev og:description: A hands-on benchmark comparing Qwen 3.6 35B (A3B) with and without Multi-Token Prediction (MTP) enabled, running locally on an Apple M5 Max via LM Studio.... og:url: https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx og:image: https://api.daily.dev/og/posts/GmvU4eWLX.png og:image:alt: QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It? 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. # QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It? **[Execute Automation](https://daily.dev/sources/executeautomation)** · 12 min read · 1 upvotes · 0 comments ## Summary A hands-on benchmark comparing Qwen 3.6 35B (A3B) with and without Multi-Token Prediction (MTP) enabled, running locally on an Apple M5 Max via LM Studio. Using a PI coding agent as the workload, the MTP model achieved ~109 t/s in round one but dropped to ~85 t/s in round two (average 97.7 t/s), while the non-MTP model held steadier at ~89-90 t/s across both rounds. MTP wins on average speed (~8.8% faster) and produces more verbose output, but its draft token acceptance rate degrades over longer sessions as context grows. Claude Desktop was used to analyze and visualize the raw stats. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Up to 580tps\! New Speed Record of Qwen3.5-397B-A17B on GPU for Agentic Workloads with TokenSpeed – PyTorch](https://daily.dev/posts/up-to-580tps-new-speed-record-of-qwen3-5-397b-a17b-on-gpu-for-agentic-workloads-with-tokenspeed-p-imoe4kosk) · PyTorch · 0 upvotes · 0 comments --- Tags: [#ai-inference](https://daily.dev/tags/ai-inference), [#lm-studio](https://daily.dev/tags/lm-studio), [#local-ai](https://daily.dev/tags/local-ai), [#qwen](https://daily.dev/tags/qwen) [View this post on daily.dev](https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx) ```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":"QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It?","url":"https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx"},"datePublished":"2026-06-10T10:04:45.906Z","dateModified":"2026-06-18T02:15:44.956Z","description":"A hands-on benchmark comparing Qwen 3.6 35B (A3B) with and without Multi-Token Prediction (MTP) enabled, running locally on an Apple M5 Max via LM Studio....","image":"https://i.ytimg.com/vi/DhGUi4CtJRg/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/DhGUi4CtJRg/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Execute Automation","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":"Execute Automation","logo":"https://media.daily.dev/image/upload/s--Uh-EE9Cf--/f_auto,q_auto/v1780213688/logos/executeautomation?_a=BAMAMiWQ0","url":"https://daily.dev/sources/executeautomation"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/qwen3-6-mtp-vs-non-mtp-on-m5-max-i-hit-113-t-s-but-which-one-did-it--gmvu4ewlx","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-inference,lm-studio,local-ai,qwen","timeRequired":"PT12M","video":{"@type":"VideoObject","name":"QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It?","description":"A hands-on benchmark comparing Qwen 3.6 35B (A3B) with and without Multi-Token Prediction (MTP) enabled, running locally on an Apple M5 Max via LM Studio....","thumbnailUrl":"https://i.ytimg.com/vi/DhGUi4CtJRg/sddefault.jpg","uploadDate":"2026-06-10T10:04:45.906Z","duration":"PT12M","url":"https://api.daily.dev/r/GmvU4eWLX","embedUrl":"https://www.youtube.com/embed/DhGUi4CtJRg"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Execute Automation","item":"https://daily.dev/sources/executeautomation"},{"@type":"ListItem","position":3,"name":"QWEN3.6 MTP vs Non-MTP on M5 Max — I Hit 113 t/s, But Which One Did It?"}]} ```