--- title: Kimi K3 is now available via multiple APIs and inference... description: Moonshot's Kimi K3, a 2.8 trillion parameter mixture-of-experts model with a 1 million token context window and native vision, is now publicly available with... canonical: https://daily.dev/posts/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Kimi K3 is now available via multiple APIs and inference providers | daily.dev og:description: Moonshot's Kimi K3, a 2.8 trillion parameter mixture-of-experts model with a 1 million token context window and native vision, is now publicly available with... og:url: https://daily.dev/posts/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz og:image: https://api.daily.dev/og/posts/yJd0gsnJZ.png og:image:alt: Kimi K3 is now available via multiple APIs and inference providers 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 is now available via multiple APIs and inference providers **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 0 comments ## Summary Moonshot's Kimi K3, a 2.8 trillion parameter mixture-of-experts model with a 1 million token context window and native vision, is now publicly available with weights, a technical report, and same-day inference support. It ranks first among open models on Artificial Analysis's Intelligence Index and fourth overall across 186 models. Inference is available via Modal (with a custom DFlash speculator boosting throughput to 100+ tokens/sec per GPU), Fireworks AI (including LoRA fine-tuning), and vLLM with a canonical deployment guide. Key architectural features include Kimi Delta Attention for long-context efficiency, Attention Residuals, and MXFP4/MXFP8 quantization-aware training for broader hardware compatibility. Distilled smaller versions are expected to follow. ## Content Moonshot's Kimi K3 is out in the open - weights, technical report, and same-day inference support across several platforms. The model is a 2.8 trillion parameter mixture-of-experts with a 1 million token context window and native vision. It currently ranks first among open models on Artificial Analysis's Intelligence Index, and fourth overall across 186 models. Those are strong numbers, though the sheer size means you won't be running this locally anytime soon. ## Where to run it **Modal** partnered with Moonshot and vLLM for day-zero support. They offer it through a Shared API with token-based pricing and as a dedicated Auto Endpoint. They also trained a custom DFlash speculator tuned to K3's architecture, which pushes decode throughput from around 50 tokens/sec to over 100 tokens/sec per GPU, with per-user peaks above 200 tokens/sec. **Fireworks AI** has it available for both inference and fine-tuning. LoRA adapter training on a frontier model this size is genuinely useful if you want to customize behavior without running your own infrastructure. **vLLM** published a canonical deployment guide covering architecture, kernels, and production flags. K3 was supported from day zero. There's also at least one US-hosted provider that's reportedly fast for a model this size - worth trying if you just want to poke at it without setting anything up. ## What's under the hood A few architectural details worth knowing: - **Kimi Delta Attention** handles long-context efficiently - **Attention Residuals** improve scaling efficiency - Weights use MXFP4, activations use MXFP8, both from quantization-aware training - this is what makes it deployable on reasonably broad hardware ## What to expect going forward Pricing will likely drop as providers optimize further. Distilled versions are probably coming too - smaller models trained on K3's outputs that can run locally while retaining a decent chunk of the capability. That's been the pattern with every large open release, and there's no reason K3 would be different. --- Tags: [#llm](https://daily.dev/tags/llm), [#vllm](https://daily.dev/tags/vllm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts), [#kimi-k3](https://daily.dev/tags/kimi-k3) [View this post on daily.dev](https://daily.dev/posts/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz) ```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 is now available via multiple APIs and inference providers","url":"https://daily.dev/posts/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz"},"datePublished":"2026-07-27T19:56:17.347Z","dateModified":"2026-07-27T19:56:55.926Z","description":"Moonshot's Kimi K3, a 2.8 trillion parameter mixture-of-experts model with a 1 million token context window and native vision, is now publicly available with...","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/kimi-k3-is-now-available-via-multiple-apis-and-inference-providers-yjd0gsnjz","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,vllm,mixture-of-experts,kimi-k3","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":"Kimi K3 is now available via multiple APIs and inference providers"}]} ```