--- title: Intology's Locus system post-trained Qwen3 1.7B to beat... description: Intology's automated research system, Locus, post-trained Qwen3 base models and outperformed the official Qwen3 1.7B Instruct release on PostTrainBench,... canonical: https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Intology's Locus system post-trained Qwen3 1.7B to beat the official instruct release | daily.dev og:description: Intology's automated research system, Locus, post-trained Qwen3 base models and outperformed the official Qwen3 1.7B Instruct release on PostTrainBench,... og:url: https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m og:image: https://api.daily.dev/og/posts/l7CDlhi5M.png og:image:alt: Intology's Locus system post-trained Qwen3 1.7B to beat the official instruct release 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. # Intology's Locus system post-trained Qwen3 1.7B to beat the official instruct release **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 1 comments ## Summary Intology's automated research system, Locus, post-trained Qwen3 base models and outperformed the official Qwen3 1.7B Instruct release on PostTrainBench, achieving state-of-the-art results. The key implication: an automated system matched and exceeded the post-training work done manually by Qwen's team. ## Content Intology's automated research system, Locus, has post-trained a model that beats the official human-tuned Qwen3-1.7B Instruct release: 51.6% vs 49.4% on PostTrainBench, the benchmark for evaluating AI agents that post-train other models. The standard PostTrainBench setup gives each agent one H100 and 10 hours. Intology removed that constraint and ran Locus for 100 hours across a cluster, raising the total compute budget from 70 H100-hours to 4,500. They're calling this extended setup PostTrainBench+. What Locus is actually doing during those 100 hours is the interesting part: allocating compute, running parallel experiments, reading evaluation results, dropping weak approaches, and doubling down on promising ones. That's the core loop of ML research, and it's running without a human in it. There are real caveats. One run per setting is a significant limitation - it's hard to draw strong conclusions from a single trial. And expanding the compute budget by 64x is a meaningful change to the evaluation conditions, not just a minor tweak. Still, the result makes a reasonable case that automated research systems should be evaluated at longer timescales, where the compounding effect of many sequential decisions actually shows up. A 10-hour window might not be long enough to see what these systems can do. ## Community discussion Top comments from developers on daily.dev. **@amitgajare** · 0 upvotes > Explain better --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#qwen](https://daily.dev/tags/qwen) [View this post on daily.dev](https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m) ```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":"Intology's Locus system post-trained Qwen3 1.7B to beat the official instruct release","url":"https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m"},"datePublished":"2026-08-03T18:09:03.734Z","dateModified":"2026-08-03T19:03:27.026Z","description":"Intology's automated research system, Locus, post-trained Qwen3 base models and outperformed the official Qwen3 1.7B Instruct release on PostTrainBench,...","image":"https://pbs.twimg.com/media/HO0QfCZWwAAFc3n.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HO0QfCZWwAAFc3n.jpg","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":1,"discussionUrl":"https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"machine-learning,llm,qwen","timeRequired":"PT2M"} {"@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":"Intology's Locus system post-trained Qwen3 1.7B to beat the official instruct release"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/intology-s-locus-system-post-trained-qwen3-1-7b-to-beat-the-official-instruct-release-l7cdlhi5m","comment":[{"@type":"Comment","text":"Explain better","datePublished":"2026-08-03T18:26:17.490Z","url":"https://daily.dev/posts/l7CDlhi5M#c-Qq1Mct6VA","author":{"@type":"Person","name":"Amit Gajare","url":"https://daily.dev/amitgajare","image":"https://media.daily.dev/image/upload/s--7lr6XGRA--/f_auto/v1735367756/avatars/avatar_UU1stgGJldM9pQP0QluF0"}}]} ```