--- title: State of Open Models: Summer 2026 Observations | daily.dev description: Hugging Face's biannual state-of-open-models report covers January-August 2026. Chinese labs (Moonshot, MiniMax, Z.ai, DeepSeek) are shipping frontier-scale... canonical: https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: State of Open Models: Summer 2026 Observations | daily.dev og:description: Hugging Face's biannual state-of-open-models report covers January-August 2026. Chinese labs (Moonshot, MiniMax, Z.ai, DeepSeek) are shipping frontier-scale... og:url: https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq og:image: https://api.daily.dev/og/posts/VBG4bEkbq.png og:image:alt: State of Open Models: Summer 2026 Observations 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. # State of Open Models: Summer 2026 Observations **[Hugging Face](https://daily.dev/sources/huggingface)** · 14 min read · 0 upvotes · 0 comments ## Summary Hugging Face's biannual state-of-open-models report covers January-August 2026. Chinese labs (Moonshot, MiniMax, Z.ai, DeepSeek) are shipping frontier-scale open models far larger than US labs, often skipping small-model releases entirely, while American open-source activity has shifted from model labs like Meta and Google toward hardware vendors NVIDIA and AMD. Downloads and likes measure different things: likes track excitement about new frontier releases while downloads track long-term infrastructure dependence, with old small models like all-MiniLM-L6-v2 dominating raw download counts. Chinese frontier labs license their largest models under permissive Apache 2.0/MIT terms rather than restrictive licenses, suggesting monetization comes from API/cloud/hardware rather than licensing. Qwen has become the ecosystem's base model with over 151,000 derivatives, far ahead of Llama, driven by consistent releases, broad size coverage, and permissive licensing. llama.cpp and GGUF quantization now let trillion-parameter models run on consumer hardware, and GGUF-related repository growth is outpacing the platform average. A new agent-usage dataset shows coding agents (Claude Code, Codex) generating a volatile and fast-growing share of Hub traffic, and the report also discloses a security incident where an autonomous agent conducted an intrusion, analyzed afterward using an open model since closed frontier models refused due to safety guardrails. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why do frontier open models get way fewer downloads than small models like all-MiniLM-L6-v2 on Hugging Face? Downloads and likes measure different behaviors: likes spike around a frontier model's release as a signal of excitement, while downloads accrue over years to small, stable models wired into recurring pipelines. All-MiniLM-L6-v2 was pulled 1.55 billion times in seven months versus 5,156 likes, while Kimi-K3 got about 60 downloads per like, showing attention and adoption diverge sharply. _Anyone weighing hype against real adoption for a model can track ecosystem data like this on daily.dev._ ### How are trillion-parameter open models like Kimi-K3 actually run locally by developers? They run through llama.cpp and GGUF quantization, which lets mixture-of-experts models spread across a few consumer machines instead of requiring datacenter hardware. GGUF builds of Kimi-K3 at roughly 2.8 trillion parameters and DeepSeek-V4-Flash at roughly 284B parameters exist, and repositories declaring the gguf library grew 464% over seven months, far outpacing the 16% growth of transformers. _Developers piecing together local inference workflows for huge models can follow tooling shifts like this on daily.dev._ ### Why has Qwen become more widely adopted than Llama for fine-tuning and local deployment? Qwen has 151,448 derivatives on Hugging Face, 4.7 times Llama's repository count, driven by consistent release cadence, coverage across model sizes from under 1B to 2.4 trillion parameters, and permissive Apache 2.0 licensing. Qwen also gets 39.6 million GGUF downloads a month versus Llama's 7.5 million, despite Llama having a comparable number of GGUF repositories available. _Teams deciding which base model family to standardize on can weigh adoption signals like these via daily.dev._ ## Similar posts on daily.dev - [State of Open Source on Hugging Face: Spring 2026](https://daily.dev/posts/state-of-open-source-on-hugging-face-spring-2026-fuzipfbpa) · Hugging Face · 2 upvotes · 0 comments - [The state of open source AI models in 2025](https://daily.dev/posts/the-state-of-open-source-ai-models-in-2025-nbwk4nxtr) · Red Hat Developer · 52 upvotes · 1 comments - [China is winning the open source AI race — but a US company still controls everything underneath](https://daily.dev/posts/china-is-winning-the-open-source-ai-race-but-a-us-company-still-controls-everything-underneath-jrc9xq6eh) · The New Stack · 2 upvotes · 0 comments - [How to Choose Between Small and Frontier Models](https://daily.dev/posts/how-to-choose-between-small-and-frontier-models-fwsu2bqxo) · Towards Data Science · 0 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#huggingface](https://daily.dev/tags/huggingface), [#qwen](https://daily.dev/tags/qwen), [#llama-cpp](https://daily.dev/tags/llama-cpp) [View this post on daily.dev](https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq) ```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":"State of Open Models: Summer 2026 Observations","url":"https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq"},"datePublished":"2026-08-14T15:14:16.169Z","dateModified":"2026-08-14T17:38:19.488Z","description":"Hugging Face's biannual state-of-open-models report covers January-August 2026. Chinese labs (Moonshot, MiniMax, Z.ai, DeepSeek) are shipping frontier-scale...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/0fe6a26c760c13d7a126aee026f3ba2d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/0fe6a26c760c13d7a126aee026f3ba2d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hugging Face","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":"Hugging Face","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/f1f55c67d81a4330acf5b90b26b0c8e1","url":"https://daily.dev/sources/huggingface"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,huggingface,qwen,llama-cpp","timeRequired":"PT14M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hugging Face","item":"https://daily.dev/sources/huggingface"},{"@type":"ListItem","position":3,"name":"State of Open Models: Summer 2026 Observations"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/state-of-open-models-summer-2026-observations-vbg4bekbq#faq","mainEntity":[{"@type":"Question","name":"Why do frontier open models get way fewer downloads than small models like all-MiniLM-L6-v2 on Hugging Face?","acceptedAnswer":{"@type":"Answer","text":"Downloads and likes measure different behaviors: likes spike around a frontier model's release as a signal of excitement, while downloads accrue over years to small, stable models wired into recurring pipelines. All-MiniLM-L6-v2 was pulled 1.55 billion times in seven months versus 5,156 likes, while Kimi-K3 got about 60 downloads per like, showing attention and adoption diverge sharply. Anyone weighing hype against real adoption for a model can track ecosystem data like this on daily.dev."}},{"@type":"Question","name":"How are trillion-parameter open models like Kimi-K3 actually run locally by developers?","acceptedAnswer":{"@type":"Answer","text":"They run through llama.cpp and GGUF quantization, which lets mixture-of-experts models spread across a few consumer machines instead of requiring datacenter hardware. GGUF builds of Kimi-K3 at roughly 2.8 trillion parameters and DeepSeek-V4-Flash at roughly 284B parameters exist, and repositories declaring the gguf library grew 464% over seven months, far outpacing the 16% growth of transformers. Developers piecing together local inference workflows for huge models can follow tooling shifts like this on daily.dev."}},{"@type":"Question","name":"Why has Qwen become more widely adopted than Llama for fine-tuning and local deployment?","acceptedAnswer":{"@type":"Answer","text":"Qwen has 151,448 derivatives on Hugging Face, 4.7 times Llama's repository count, driven by consistent release cadence, coverage across model sizes from under 1B to 2.4 trillion parameters, and permissive Apache 2.0 licensing. Qwen also gets 39.6 million GGUF downloads a month versus Llama's 7.5 million, despite Llama having a comparable number of GGUF repositories available. Teams deciding which base model family to standardize on can weigh adoption signals like these via daily.dev."}}]} ```