--- title: Tencent releases Hunyuan Hy4 Preview: 770B MoE model... description: Tencent has released Hy4 Preview, an open-weight large language model licensed under Apache 2.0. It uses a Mixture-of-Experts architecture with 770B total... canonical: https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Tencent releases Hunyuan Hy4 Preview: 770B MoE model with 1M context under Apache 2.0 | daily.dev og:description: Tencent has released Hy4 Preview, an open-weight large language model licensed under Apache 2.0. It uses a Mixture-of-Experts architecture with 770B total... og:url: https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp og:image: https://api.daily.dev/og/posts/Mwo7hWeXp.png og:image:alt: Tencent releases Hunyuan Hy4 Preview: 770B MoE model with 1M context under Apache 2.0 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. # Tencent releases Hunyuan Hy4 Preview: 770B MoE model with 1M context under Apache 2.0 **[Collections](https://daily.dev/sources/collections)** · 1 min read · 4 upvotes · 0 comments ## Summary Tencent has released Hy4 Preview, an open-weight large language model licensed under Apache 2.0. It uses a Mixture-of-Experts architecture with 770B total parameters, 49B active parameters, and a 1M token context window. Detailed evaluations and benchmarks are still pending as testing continues. ## Content Tencent has released Hy4 Preview, a new open-weight language model under the Apache 2.0 license. It's a Mixture-of-Experts architecture with 770B total parameters and 49B active at inference time, paired with a 1M-token context window. Early looks suggest it performs well on long-horizon software engineering tasks. The Apache 2.0 license means commercial use is straightforward, which puts it in a more permissive category than many recent open-weight releases. ## Questions this post answers ### What is Tencent's Hy4 Preview model and what license is it released under? Hy4 Preview is an open-weight large language model from Tencent released under the Apache 2.0 license, allowing commercial use without the restrictions found in some other open-weight licenses. It uses a Mixture-of-Experts architecture with 770 billion total parameters, 49 billion active parameters, and a 1 million token context window. Detailed benchmark evaluations have not yet been published. _Developers weighing open-weight model licenses for production use can follow releases like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 3 discussions and 8 comments across x (as of 2026-08-28). **TL;DR:** Reactions are largely intrigued by the scale (770B total, 49B active) and the true Apache 2.0 licensing, though several commenters withhold judgment until the 1M-token context is actually tested for long-range recall. **Sentiment:** 45% positive · 40% mixed · 15% skeptical **The case for** - The 49B active parameter count suggests serving costs closer to a mid-size dense model despite the huge total size. - The Apache 2.0 license is seen as genuinely permissive, without field-of-use restrictions or acceptable-use riders that other 'open weight' releases quietly impose. **The pushback** - Skepticism that the 1M-token context window will actually maintain recall at large token counts (e.g. 800k) rather than just accepting the input length. - Some see the announcement as mostly a spec-sheet claim that needs real-workload validation before it can be trusted. **By community** - x (mixed): Excitement about the model's scale and license is tempered by calls to validate real-world long-context performance before drawing conclusions. **Open questions** - Does the 1M-token context window actually preserve recall at high token counts, or does performance degrade well before the stated ceiling? - How does the model perform against other MoE models in real workloads rather than benchmarks? **Highlights** > @testingcatalog Apache 2.0 is the actual news item here — no field-of-use restrictions, no acceptable-use rider, commercial deployment without asking anyone. Most "open weight" releases from large labs come with a custom license that quietly isn't; this one doesn't. > — [ricci\_nov on x](https://x.com/ricci_nov/status/2093329733832376606) > @testingcatalog 49B active out of 770B is the number that matters here, that is a serving cost close to a mid size dense model. Apache 2.0 on top makes it the interesting one to test this week. Will see if the 1M context actually holds recall past a few hundred k. > — [Flextor97 on x](https://x.com/Flextor97/status/2093260858705252683) > @omarsar0 1M context is the spec sheet number. the real test for long-horizon software work is whether recall holds at token 800k, not whether the window accepts the input, most models advertising huge windows start dropping details well before the ceiling. > — [vsaietta on x · 1 points](https://x.com/vsaietta/status/2093347108929536322) **Source threads** - [x](https://x.com/scaling01/status/2093309623780421875) · 0 points · 0 comments - [x](https://x.com/testingcatalog/status/2093229582992154830) · 1 points · 7 comments - [x](https://x.com/omarsar0/status/2093342979276648539) · 0 points · 1 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts) [View this post on daily.dev](https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp) ```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":"Tencent releases Hunyuan Hy4 Preview: 770B MoE model with 1M context under Apache 2.0","url":"https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp"},"datePublished":"2026-08-28T12:08:29.024Z","dateModified":"2026-08-28T20:21:35.618Z","description":"Tencent has released Hy4 Preview, an open-weight large language model licensed under Apache 2.0. It uses a Mixture-of-Experts architecture with 770B total...","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/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,mixture-of-experts","timeRequired":"PT1M"} {"@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":"Tencent releases Hunyuan Hy4 Preview: 770B MoE model with 1M context under Apache 2.0"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/tencent-releases-hunyuan-hy4-preview-770b-moe-model-with-1m-context-under-apache-2-0-mwo7hwexp#faq","mainEntity":[{"@type":"Question","name":"What is Tencent's Hy4 Preview model and what license is it released under?","acceptedAnswer":{"@type":"Answer","text":"Hy4 Preview is an open-weight large language model from Tencent released under the Apache 2.0 license, allowing commercial use without the restrictions found in some other open-weight licenses. It uses a Mixture-of-Experts architecture with 770 billion total parameters, 49 billion active parameters, and a 1 million token context window. Detailed benchmark evaluations have not yet been published. Developers weighing open-weight model licenses for production use can follow releases like this on daily.dev."}}]} ```