--- title: RedNote releases dots3-note Preview, an open model built... description: RedNote's dotsstudioai team released dots3-note Preview, an open-weight model under Apache 2.0 aimed at long-horizon agent tasks, with day-0 support in vLLM.... canonical: https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: RedNote releases dots3-note Preview, an open model built for long-running agents | daily.dev og:description: RedNote's dotsstudioai team released dots3-note Preview, an open-weight model under Apache 2.0 aimed at long-horizon agent tasks, with day-0 support in vLLM.... og:url: https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw og:image: https://api.daily.dev/og/posts/SnCLEp8iw.png og:image:alt: RedNote releases dots3-note Preview, an open model built for long-running agents 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. # RedNote releases dots3-note Preview, an open model built for long-running agents **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 0 comments ## Summary RedNote's dotsstudioai team released dots3-note Preview, an open-weight model under Apache 2.0 aimed at long-horizon agent tasks, with day-0 support in vLLM. It has 280B total parameters (16B active, MoE), a 512K context window, and handles text, images, audio, and video. The model uses a reinforcement learning technique called TEMPO, where it plays both actor and critic roles, periodically pausing to check whether it's actually progressing toward a goal or just drifting. The aim is to address the tendency of long-running agents to lose track of shifting goals or stale assumptions over extended tasks. ## Content There's a new open-weight multimodal model out from rednote's dots studio, and the interesting part isn't the size (though the size is notable too). It's called **dots3-note Preview**, and it's aimed at a problem that's genuinely hard: keeping an AI agent useful when a task runs for hours, or even days, while the world underneath it keeps shifting. Anyone who's run a long-horizon agent knows the pain here. Goals change mid-task. New information shows up. Plans that made sense an hour ago stop making sense. Most agents just plow ahead anyway, burning compute on a stale objective. ## The basics - 280B total parameters, 16B active (mixture-of-experts) - 512K context window - Handles text, vision, and speech/audio - Reasoning, coding, and tool use - Released under Apache 2.0, with day-0 support in vLLM So it's a big model that's cheap to run relative to its size, and you can read images, audio, and video into it. That part is table stakes these days. What's actually worth paying attention to is how it handles time. ## The interesting bit: TEMPO The model introduces something called TEMPO, a reinforcement learning technique that lets the model stop mid-task and grade its own progress. Instead of grinding forward blindly, the same model periodically switches roles: it acts as the **actor** working the problem, then flips to **critic** and asks something close to ## Questions this post answers ### What is dots3-note Preview and what are its specs? Dots3-note Preview is an open-weight model from RedNote's dotsstudioai team, released under Apache 2.0 with day-0 support in vLLM. It has 280B total parameters with 16B active (mixture-of-experts architecture), a 512K context window, and can process text, images, audio, and video. It's trained specifically for reasoning, coding, tool use, and long-horizon agent work. _Track new open-weight agent model releases like this one as they land on daily.dev._ ### How does TEMPO help AI agents avoid drifting on long tasks? TEMPO is a reinforcement learning technique used in dots3-note Preview where the same model periodically pauses to check its own progress, alternating between an actor role that pushes the task forward and a critic role that evaluates whether the agent is actually closer to the goal or just spinning its wheels. This critique feeds back into the actor's next steps to correct drift on long-running, messy tasks. _Developers evaluating self-correcting agent techniques can follow this space on daily.dev._ --- Tags: [#open-source](https://daily.dev/tags/open-source), [#ai-agents](https://daily.dev/tags/ai-agents), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning), [#vllm](https://daily.dev/tags/vllm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts) [View this post on daily.dev](https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw) ```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":"RedNote releases dots3-note Preview, an open model built for long-running agents","url":"https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw"},"datePublished":"2026-08-19T16:15:59.079Z","dateModified":"2026-08-24T13:55:13.294Z","description":"RedNote's dotsstudioai team released dots3-note Preview, an open-weight model under Apache 2.0 aimed at long-horizon agent tasks, with day-0 support in vLLM....","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":0,"discussionUrl":"https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,ai-agents,reinforcement-learning,vllm,mixture-of-experts","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":"RedNote releases dots3-note Preview, an open model built for long-running agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/rednote-releases-dots3-note-preview-an-open-model-built-for-long-running-agents-snclep8iw#faq","mainEntity":[{"@type":"Question","name":"What is dots3-note Preview and what are its specs?","acceptedAnswer":{"@type":"Answer","text":"Dots3-note Preview is an open-weight model from RedNote's dotsstudioai team, released under Apache 2.0 with day-0 support in vLLM. It has 280B total parameters with 16B active (mixture-of-experts architecture), a 512K context window, and can process text, images, audio, and video. It's trained specifically for reasoning, coding, tool use, and long-horizon agent work. Track new open-weight agent model releases like this one as they land on daily.dev."}},{"@type":"Question","name":"How does TEMPO help AI agents avoid drifting on long tasks?","acceptedAnswer":{"@type":"Answer","text":"TEMPO is a reinforcement learning technique used in dots3-note Preview where the same model periodically pauses to check its own progress, alternating between an actor role that pushes the task forward and a critic role that evaluates whether the agent is actually closer to the goal or just spinning its wheels. This critique feeds back into the actor's next steps to correct drift on long-running, messy tasks. Developers evaluating self-correcting agent techniques can follow this space on daily.dev."}}]} ```