--- title: Kimi K3 just parameter mogged every open-weight model… description: Moonshot AI released Kimi K3, a massive open-weight model with 2.8 trillion parameters, a 1 million token context window, and a mixture-of-experts architecture... canonical: https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Kimi K3 just parameter mogged every open-weight model… | daily.dev og:description: Moonshot AI released Kimi K3, a massive open-weight model with 2.8 trillion parameters, a 1 million token context window, and a mixture-of-experts architecture... og:url: https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo og:image: https://api.daily.dev/og/posts/Qza9G1qyO.png og:image:alt: Kimi K3 just parameter mogged every open-weight model… 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. # Kimi K3 just parameter mogged every open-weight model… **[Fireship](https://daily.dev/sources/fireship)** · 5 min read · 16 upvotes · 1 comments ## Summary Moonshot AI released Kimi K3, a massive open-weight model with 2.8 trillion parameters, a 1 million token context window, and a mixture-of-experts architecture activating 16 of 896 experts per token. It ranks #1 on the Frontend Code Arena benchmark (1,679 ELO), beating Claude and GPT-5 equivalents on some coding benchmarks, though Moonshot acknowledges it still trails on overall benchmarks like Humanity's Last Exam by ~10 points. A 51% hallucination rate and verbose token output are notable weaknesses. The release sparked geopolitical debate, with US officials considering entity-listing Chinese AI labs and OpenAI arguing open weights are 'decelerationist.' Weights are expected to drop July 27th but require data-center-grade hardware to self-host. Alibaba also released Qwen 3.8 with 2.4 trillion parameters simultaneously. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How many parameters and experts does Kimi K3 have? Kimi K3 is a mixture-of-experts model with 2.8 trillion total parameters, 896 total experts, and only 16 experts activated per token. It has a 1 million token context window and is optimized for long-horizon reasoning and coding tasks, making it roughly 2.5 times more scaling-efficient than its predecessor K2. _Track major open-weight model releases like this one as they reshape which coding tools are worth adopting, via daily.dev._ ### How does Kimi K3 compare to Claude and GPT on coding benchmarks? Kimi K3 ranks number one on the frontend code arena with a 1,679 ELO, ahead of Claude and GPT-5.6, and lands in the top three on the artificial analysis intelligence index. However, many of these results used Moonshot's own Kimi code harness while competitors ran different harnesses, and Moonshot itself admits K3 still trails on tests like Humanity's Last Exam by about 10 points, with a 51% hallucination rate reported by artificial analysis. _Compare coding model benchmarks carefully before switching stacks, something daily.dev readers weigh constantly._ ### When are the Kimi K3 model weights being released and can I self-host it? Weights are expected to be released around July 27th, and self-hosting is technically possible since the model is open-weight, but running a 2.8 trillion parameter model requires a massive array of data-center-caliber GPUs, not a consumer gaming GPU. Hosted paid plans sold out shortly after launch due to high demand. _Developers evaluating whether to self-host huge open models can weigh options like this through daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@harshitmanepally** · 2 upvotes >  --- Tags: [#ai](https://daily.dev/tags/ai), [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts), [#kimi](https://daily.dev/tags/kimi) [View this post on daily.dev](https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo) ```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":"Kimi K3 just parameter mogged every open-weight model…","url":"https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo"},"datePublished":"2026-07-22T17:48:45.833Z","dateModified":"2026-09-14T06:14:17.504Z","description":"Moonshot AI released Kimi K3, a massive open-weight model with 2.8 trillion parameters, a 1 million token context window, and a mixture-of-experts architecture...","image":"https://i.ytimg.com/vi/YP73B9D20V4/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/YP73B9D20V4/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Fireship","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":"Fireship","logo":"https://media.daily.dev/image/upload/s--75ndqrkr--/f_auto,t_logo/v1702882094/logos/fireship.jpg","url":"https://daily.dev/sources/fireship"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":16},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai,open-source,llm,mixture-of-experts,kimi","timeRequired":"PT5M","video":{"@type":"VideoObject","name":"Kimi K3 just parameter mogged every open-weight model…","description":"Moonshot AI released Kimi K3, a massive open-weight model with 2.8 trillion parameters, a 1 million token context window, and a mixture-of-experts architecture...","thumbnailUrl":"https://i.ytimg.com/vi/YP73B9D20V4/sddefault.jpg","uploadDate":"2026-07-22T17:48:45.833Z","duration":"PT5M","url":"https://api.daily.dev/r/Qza9G1qyO","embedUrl":"https://www.youtube.com/embed/YP73B9D20V4"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Fireship","item":"https://daily.dev/sources/fireship"},{"@type":"ListItem","position":3,"name":"Kimi K3 just parameter mogged every open-weight model…"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo","comment":[{"@type":"Comment","text":"","datePublished":"2026-07-26T07:53:25.731Z","url":"https://daily.dev/posts/Qza9G1qyO#c-neX2N8erT","author":{"@type":"Person","name":"HARSHIT MANEPALLY","url":"https://daily.dev/harshitmanepally","image":"https://media.daily.dev/image/upload/s--O0TOmw4y--/f_auto/v1715772965/public/noProfile"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/kimi-k3-just-parameter-mogged-every-open-weight-model--qza9g1qyo#faq","mainEntity":[{"@type":"Question","name":"How many parameters and experts does Kimi K3 have?","acceptedAnswer":{"@type":"Answer","text":"Kimi K3 is a mixture-of-experts model with 2.8 trillion total parameters, 896 total experts, and only 16 experts activated per token. It has a 1 million token context window and is optimized for long-horizon reasoning and coding tasks, making it roughly 2.5 times more scaling-efficient than its predecessor K2. Track major open-weight model releases like this one as they reshape which coding tools are worth adopting, via daily.dev."}},{"@type":"Question","name":"How does Kimi K3 compare to Claude and GPT on coding benchmarks?","acceptedAnswer":{"@type":"Answer","text":"Kimi K3 ranks number one on the frontend code arena with a 1,679 ELO, ahead of Claude and GPT-5.6, and lands in the top three on the artificial analysis intelligence index. However, many of these results used Moonshot's own Kimi code harness while competitors ran different harnesses, and Moonshot itself admits K3 still trails on tests like Humanity's Last Exam by about 10 points, with a 51% hallucination rate reported by artificial analysis. Compare coding model benchmarks carefully before switching stacks, something daily.dev readers weigh constantly."}},{"@type":"Question","name":"When are the Kimi K3 model weights being released and can I self-host it?","acceptedAnswer":{"@type":"Answer","text":"Weights are expected to be released around July 27th, and self-hosting is technically possible since the model is open-weight, but running a 2.8 trillion parameter model requires a massive array of data-center-caliber GPUs, not a consumer gaming GPU. Hosted paid plans sold out shortly after launch due to high demand. Developers evaluating whether to self-host huge open models can weigh options like this through daily.dev."}}]} ```