The ATOM Project - American Truly Open Models
ATOM: American Truly Open Models
The ATOM Project
Towards fully open models for US research & industry
August 4, 2025Nathan LambertData & MethodsSign Your Support
Reinvigorating AI research in the U.S. by building leading, open models at home
America's AI leadership was built by being the global hub and leading producer of open AI research, research which led directly to innovations like the Transformer architecture, ChatGPT, and the latest innovations in reasoning models and agents. America is poised to lose this leadership to China, in a period of geopolitical uncertainty and rising tensions between these two nations. America's best AI models have become more closed and restricted, while Chinese models have become more open, capturing substantial market share from businesses and researchers in the U.S. and abroad.
Open language models
Open Language Models
AI models whose weights, training code, and sometimes training data are freely available for download, modification, and use by researchers and developers worldwide without restrictions. Generally they are split into two types, open-weight models with just the weights for inference or finetuning available and open source models that have all of that along with training data and code.
are becoming the foundation of AI research and the most important tool in securing this leadership. America has lost its lead in open models – both in performance and adoption – and is on pace to fall further behind. The United States must lead AI research globally, and we must invest in making the tools our researchers need to do their job here in America: a suite of leading, open foundation models that can re-establish the strength of the research ecosystem.
1. Global Model Momentum
Open models from China are set to overtake the US in the near future, as downloads surge in usage across HuggingFace
HuggingFace
The leading platform for hosting and distributing open AI models, where researchers share model weights and collaborate on AI development.
. The American ecosystem has stalled in growth, and the "flip" is is happening right now.
Models Worldwide
Cumulative Downloads, 2023- present
USA
China
EU
The ATOM Project (updated 03/2026)
source: huggingface
Recommendation: To regain global leadership in open source AI, America needs to maintain multiple labs focused on training open models with 10,000+ leading-edge GPUs. The PRC currently has at least five labs producing and releasing open models at or beyond the capabilities of the best U.S. open model. Regaining open source leadership is necessary to drive research into fundamental AI advances, to maximize U.S. AI market share, and to secure the U.S. AI stack.
Sign Your Support
We need your support to build American Truly Open Models and to ensure that the United States maintains its lead in AI. Sign here to lend your support to the ATOM Project:
Full Name
Email Address
Position/Title
Send me project updates by email (required)
Sign On to Support
By signing, you agree to have your name and title displayed publicly as a supporter. We will not share your personal information with third parties without your consent.
Notable Signatories
Clement DelangueCEO of Hugging Face
Jeremy HowardCo-founder of Fast.ai & Answer.ai
Oleksii KuchaievDirector of Applied Research at Nvidia
Ross TaylorCEO of General Reasoning
Sebastian RaschkaAuthor of Build A Large Language Model (From Scratch)
Soumith ChintalaCo-founder of PyTorch
Miles BrundageFormer Head of Policy Research at OpenAI
Ali FarhadiCEO of Ai2
Sergey LevineProfessor at U.C. Berkeley, Co-founder of Physical Intelligence
Bill GurleyGeneral Partner at Benchmark
Vincent WeisserCEO of Prime Intellect
Dylan PatelFounder and CEO of SemiAnalysis
Christopher D. ManningProfessor at Stanford University
Andrew TraskFounder of OpenMined
Percy LiangProfessor at Stanford University
Clement DelangueCEO of Hugging Face
Jeremy HowardCo-founder of Fast.ai & Answer.ai
Oleksii KuchaievDirector of Applied Research at Nvidia
Ross TaylorCEO of General Reasoning
Sebastian RaschkaAuthor of Build A Large Language Model (From Scratch)
Soumith ChintalaCo-founder of PyTorch
Miles BrundageFormer Head of Policy Research at OpenAI
Ali FarhadiCEO of Ai2
Sergey LevineProfessor at U.C. Berkeley, Co-founder of Physical Intelligence
Bill GurleyGeneral Partner at Benchmark
Vincent WeisserCEO of Prime Intellect
Dylan PatelFounder and CEO of SemiAnalysis
Christopher D. ManningProfessor at Stanford University
Andrew TraskFounder of OpenMined
Percy LiangProfessor at Stanford University
Overview
Open language model weights
Open Language Model Weights
Open language model weights are the learned parameters of AI models that are freely available for download and use. Unlike closed models where the weights are proprietary, open weights allow researchers and developers to study, modify, and build upon existing AI systems.
and data are the core currency of recent AI research – these are the artifacts that people use to come up with new architectures, training paradigms, or tools that will lead to the next paradigms in AI to rival The Transformer
The Transformer
The Transformer is a neural network architecture invented by Google that uses attention mechanisms to process sequences of data. It's the foundation for all modern language models like ChatGPT, Gemini, and Claude.
or Inference-time Scaling
Inference-time Scaling
A technique where AI models spend more computational resources during inference (when generating answers) to produce better results. This allows smaller models to achieve performance comparable to larger ones by "thinking" longer.
. These research advances provide continued progress on existing products or form the basis for new technology companies. At the same time, open language models create potential for a broader suite of AI offerings by allowing anyone to build and modify AI how they see fit, without their data being sent through the cloud to a few, closed model providers.
Open language models are crucial for long-term competition within American industry. Today, substantial innovation is happening inside of large, closed AI laboratories, but these groups can only cover so many of the potential ideas. These companies spend the vast majority of their resources focusing on the next model they need to train, where the broader, open research community focuses on innovations that’ll be transformative in 2, 5, 10, or more years. The most progress in building useful, intelligent AI systems will come when the most people can participate in improving today's state-of-the-art, rather than the select few at certain companies.
The open AI ecosystem (regarding the models, not to be confused with the company OpenAI) has historically been defined by many parties participating. The United States emerged as a hub of the deep learning revolution via close collaboration between leading technology companies and academic institutions. Following ChatGPT, there have been countless contributions from around the globe. This distribution of impact on research has been collapsing towards clear Chinese leadership due to their commitment to open innovation, while a large proportion of leading scientists working in the United States have joined closed research organizations.
The playbook that led Google to invent and share the Transformer
The Transformer
The Transformer is a neural network architecture invented by Google that uses attention mechanisms to process sequences of data. It's the foundation for all modern language models like ChatGPT, Gemini, and Claude.
– the defining language model architecture of which all leading models such as ChatGPT, Gemini, or Claude are derived from – is now the standard mode of operation for Chinese companies, but it is increasingly neglected by American companies.
The impact of China’s models and research are growing because the institutions focused on open models have access to substantial compute resources for training – e.g. some have formed a close relationship between leading AI training laboratories and academic institutions. Until the United States and its partners directly invest in training more, higher performance open models and sharing the processes to do so, its pace of progress in AI research will lag behind.
To train open models at the frontier of performance, a developer currently needs a high concentration of capital and talent. We estimate that to lead in open model development, the United States needs to invest in multiple clusters of 10,000+ H100 level GPUs to create an ecosystem of fully open language models that are designed to enable a resurgence in Western AI research. Stacking large investments such as this into a few focused efforts will help them to learn from each other and make progress across a range of challenges quickly and robustly. Splitting such an investment in AI training into smaller, widespread projects will not be sufficient to build leading models due to a lack of compute concentration. Along the way we need to build models of various sizes that can enable applications of AI at every scale from local or edge devices all the way to high performance cloud computing.
American model builders in 2025
Note: This list was added on November 23rd, 2025 to provide more context for readers of the ATOM Project. The rest of the text still is evergreen and accurate.
The U.S. has a comparable number of labs releasing high quality models as China (which is ~20 labs), but many American labs are releasing smaller models with more restrictive licenses , resulting in a far more muted impact. This list includes each notable recent model from them. All of these are pretrained by U.S. companies. To start, the clear players, listed alphabetically:
- Ai2 — Olmo : Open-source leader, to date fully open, smaller dense models mostly. E.g. Olmo 3 32B Think, the best fully open reasoning model (and best fully open LM ever made). Always developer friendly Apache 2.0 licenses.
- Arcee / Datology / Prime Intellect — Trinity : Newer startup fully committed to this. Range of MoE models, Apache 2.0 coming soon.
- Google — Gemma : One of the most consistent players. To date models have been smaller and dense as well. E.g. VaultGemma 1B. Often solid custom licenses, sometimes minor downstream use restrictions.
- HuggingFace — SmolLM : Tiny models, fully open, great community. E.g. SmolLM3-3B.
- IBM — Granite : Small to medium sized models, underrated, strong and consistent releases. E.g. Granite-4.0-h-1b with a new hybrid-attention architecture. Developer friendly Apache 2.0 licenses.
- Liquid AI — Liquid Foundation Models : Hybrid architecture, solid small models. E.g. LFM2-VL-3B, their first vision-language models. Very similar licenses to Apache 2.0, but with restrictions on companies making $10M per year or more.
- Microsoft — Phi (+ others) : Solid models, can go under the radar, but consistent. E.g. Phi-4-mini-flash-reasoning, a small reasoning model. Often fairly permissive licenses too, e.g. MIT.
- Moondream — Moondream Models : Solid vision models, originally on device, but dominant and consistent in their niche. E.g. Moondream3. Solid licenses and very engaged in community feedback, e.g. recently made it clear synthetic data use is okay.
- Nvidia — Nemotron : Arguably the open leader in the U.S. after Llama 4. E.g. Nemotron Nano 9B v2, a 9B model matching or surpassing Chinese models in that size range. Increasingly open licenses recently, including more open data releases.
- OpenAI — GPT OSS : A new entrant but incredibly important they're involved. E.g. gpt-oss-120b, OpenAI's first open weights language model since GPT-2. Good Apache 2.0 licenses, we hope they continue releasing more models.
- Reflection — TBD : but if you can convince someone you're worth $2B to do this, I believe you.
- ServiceNow — Apriel : Solid reasoning models and other contributions. E.g. Apriel-H1-15b-Thinker.
- Stanford University — Marin Community Models : A new entrant in fully-open models like Ai2, scaling up after releasing their first 30B base model. We hope they have strong post-training soon!
Of the above, we're watching Ai2, Nvidia, Arcee, and Reflection the closest, as the players with the most mind-share and momentum on the ground.
Unclear: Companies making fewer contributions currently but have in the past.
- Meta — Llama : The original, but priorities are changing. Crickets since the Llama 4 fiasco.
- Reka — Flash : A few solid models, no updates in a bit.
- xAI — Grok : Many promises of releasing past models, yet to be useful to the community.
- Institute for Foundation Models (IFM) : Released K2-Think & LLM 360 models. Unclear because the release was marketed as "UAE DeepSeek" but they're located within the US.
This is a list of groups making solid language models in the U.S. Other labs that would easily be included in a more "Western" ecosystem list would include the likes of Cohere (with solid models, though normally non-commercial licenses), Mistral (with smaller models, normally Apache 2.0), and AI21 labs. There are other types, such as multimodal generation models and biology-focused models that are massive breakthroughs but not listed.
The Open Model Timeline
Jan '24
DeepSeekMoE
DeepSeek
Click for details
Feb '24
Qwen 1.5
Alibaba
Click for details
Gemma
Click for details
Mar '24
xAI Grok 1
xAI
Click for details
Apr '24
Meta Llama 3
Meta
Click for details
May '24
DeepSeek-V2
DeepSeek
Click for details
Jun '24
Qwen 2
Alibaba
Click for details
NVIDIA Nemotron 4
NVIDIA
Click for details
Gemma 2
Click for details
Jul '24
Meta Llama 3.1
Meta
Click for details
Aug '24
Sep '24
Qwen 2.5
Alibaba
Click for details
DeepSeek V2.5
DeepSeek
Click for details
Oct '24
Nov '24
QwQ Preview
Alibaba
Click for details
Dec '24
DeepSeek-V3
DeepSeek
Click for details
Jan '25
DeepSeek-R1
DeepSeek
Click for details
Feb '25
Mar '25
Gemma 3
Click for details
DeepSeek-V3-0324
DeepSeek
Click for details
Apr '25
Meta Llama 4
Meta
Click for details
Qwen 3
Alibaba
Click for details
May '25
DeepSeek-R1-0528
DeepSeek
Click for details
Jun '25
MiniMax-M1
MiniMax
Click for details
ERNIE 4.5
Baidu
Click for details
Jul '25
Moonshot AI Kimi K2
Moonshot AI
Click for details
Qwen3-Coder-480B-A35B
Alibaba
Click for details
Qwen3-235B-A22B
Alibaba
Click for details
Zhipu GLM-4.5
Zhipu AI
Click for details
StepFun Step3
StepFun
Click for details
Aug '25
OpenAI GPT-OSS
OpenAI
Click for details
NVIDIA Nanotron
NVIDIA
Click for details
Seed-OSS
ByteDance
Click for details
Sep '25
Qwen Next
Qwen
Click for details
Qwen3 Omni
Qwen
Click for details
Qwen3 VL
Qwen
Click for details
GLM-4.6
Zhipu AI
Click for details
Oct '25
Ling-1T
InclusionAI
Click for details
MiniMax-M2
MiniMax
Click for details
Nov '25
Kimi-K2-Thinking
Moonshot AI
Click for details
Olmo-3-32B-Think
Allen AI
Click for details
Dec '25
DeepSeek-V3.2
DeepSeek
Click for details
Arcee Trinity
Arcee AI
Click for details
Mistral Large 3
Mistral AI
Click for details
Nemotron 3
NVIDIA
Click for details
GLM-4.7
Zhipu AI
Click for details
MiniMax-M2.1
MiniMax
Click for details
Jan '26
Kimi K2.5
Moonshot AI
Click for details
Trinity Large Preview
Arcee AI
Click for details
Feb '26
GLM-5
Z.ai
Click for details
MiniMax-M2.5
MiniMax
Click for details
Qwen3.5
Alibaba
Click for details
Mar '26
Nemotron 3 Super
NVIDIA
Click for details
Last updated March 29, 2026.
Open models as the engine for AI research and development
America's AI leadership was built by tens of thousands of our best and brightest students, academics and researchers. This process occurred over decades, but it is faltering at a crucial transition point to the new, language modeling era
Language Modeling Era
The current period in AI history where large language models trained on text data have become the dominant approach for creating general AI systems. This era began around 2017 with the Transformer architecture.
of AI research. Since the release of ChatGPT, open language models and computational resources are the most important table stakes for doing relevant and impactful research. High-quality open models and their subsequent technical reports
Technical Reports
Detailed documents that accompany AI model releases, describing the training process, architecture choices, data used, and performance results. These are crucial for scientific reproducibility and advancement.
quickly accrue thousands of citations and accolades such as best paper awards and the focus of large swaths of students. These act as foundational currencies of AI research and are crucial, achievable artifacts for the long-term American AI ecosystem.
While many direct consumers of open models are academics, this community is far from the only group that will benefit immensely from a new wave of American open models. The low cost, flexibility, and customizability of open models makes them ideal for many use cases, including many of the ways that AI stands to advance and transform businesses large and small.
If the United States does not create its own leading open models, the focus of American researchers and businesses will continue to shift abroad. The benefits of openly sharing a technology accrue to the builder in mindshare and other subtle soft power dynamics seen throughout the history of open source software. Today, these benefits are accruing elsewhere due to the intentional support of open models by many Chinese organizations. The gap in performance and adoption will only grow as the American ecosystem sees strong open models as something that is nice to have, or an afterthought, rather than a key long-term priority.
China is adopting the playbook for open innovation of language models that the United States used to create its current AI leadership, yielding rapid innovation, international adoption, and research interest. The collapse of American dominance in AI research is driven not only by the remarkable quality of the Chinese ecosystem, but also by the commitment of China to these very same Open Model Principles
Open Model Principles
The philosophy of making AI model weights, training data, and methodologies freely available to accelerate research and innovation. This includes transparency in training processes and unrestricted access to model parameters.
- the principles that American scientists used to start this AI revolution. This is reflected further in a consistent trend of Chinese open models being released with more permissive terms of use than their American counterparts.
The many leading closed research institutions in the United States are still creating world-class models – and the work they do is extraordinary. This collapse is not their fault, but closed labs make closed research, and the acceleration of AI was built on open collaboration with world-class American models as the key tool.
As researchers, our focus is on leading the research and development for the core technology defining the future, but there is also a growing list of other urgent security and policy concerns facing our nation around the lack of strong open models. To start, adoption of open models from the PRC in the US and our allies has been slow in some sectors due to worries about backdoors or poor security in generated code [1]
Security Concerns
Measuring the lack of adoption is hard, as it is tracking no signal and private communications. This sentiment has come up in many discussions we have had with companies and individuals trying to build with open models. For more information, read Interconnects.ai
. Similarly, there is concern over the outputs of these Chinese models being censored or inconsistent with everyday American values of freedom, equality, and independence [2]
Values Alignment
For more information, we recommend reading the results of SpeechMap.ai
. There are even parallels between how the PRC's national AI champions are increasingly racing to release cheap and open AI models and the PRC's historical practice of dumping state-subsidized, below-cost exports from China to undermine American competitors. With the dynamic and rapid evolution of this technology, we need to get ahead of these issues before stronger habits, cost disadvantages, or other incentives reduce the practicality of adopting American open models.
2. China's Model Momentum
Open models from China have overtaken the US, with Qwen
Qwen
An open language model series developed by Alibaba Cloud, which has gained significant adoption and represents China's growing influence in open AI.
as the most-used model family surpassing Llama in September 2025.
Catching the Llama
Cumulative Downloads, 2023-present
Llama
Qwen
Mistral
DeepSeek
OpenAI
The ATOM Project (updated 03/2026)
source: huggingface
3. Model Adoption by Region
Percentage of new finetunes and derivatives released per month by region. Each finetune is logged based on the creator of its base model. Where the U.S. and the E.U. models once dominated, China has taken a clear lead.
Global Regional Model Adoption by Month
Nov 2023 - Feb 2026
China
United States
EU
The ATOM Project (updated 03/2026)
source: huggingface
America's lost lead in open model performance
On countless benchmarks
Benchmarks
Standardized tests that measure AI model performance across various tasks like reading comprehension, math, coding, or reasoning. Examples include MMLU, HumanEval, and GSM8K.
, the leading American models have fallen behind counterparts from Chinese companies. In July 2024, American models in the form of Llama 3 had leading performance over any openly available Chinese models. Since then, a growing number of Chinese open model providers have surpassed and widened the performance gap with the leading American open models.
The leading American open models are Meta's Llama and Google's Gemma models. The Chinese open models from DeepSeek and Alibaba's Qwen
Qwen
An open language model series developed by Alibaba Cloud, which has gained significant adoption and represents China's growing influence in open AI.
have traded off positions at the frontier of capabilities ahead of their American counterparts. However, the Chinese ecosystem is expanding rapidly, with new players such as Moonshot AI (Kimi), Zhipu AI, or Tencent close behind.
We consider two popular public, aggregate benchmarks to demonstrate the state of China’s current open model dominance. These represent crowdsourced rankings, LMArena, and comprehensive intelligence rankings by blending a variety of capability benchmarks, from ArtificialAnalysis. The pace of progress on these Pareto frontiers is only part of the equation. In addition to leading, the top 10 open models on LMArena are all created by Chinese organizations. For ArtificialAnalysis rankings, the top 3 open models are of Chinese origin as of publishing on August 4th, 2025.
The isolation of Meta's Llama
Meta CEO Mark Zuckerberg has been one of the few clear advocates for the long-term imperative of America building open models. Since the release of ChatGPT, this has been manifested by Meta's Llama series of models – these had long been the definitional open models that served as the basis for research and product development in 2023 and 2024. This basis for research is established by releasing a suite of strong models across a variety of sizes. The original LLaMA family came with models of 7, 13, 32, and 65B parameters, which quickly became defaults of the research community based on convenient factors of them fitting on certain popular GPUs for finetuning or inference.
For a first instance showcasing the gap in adoption, the Qwen 1.5 family of 8 models was released shortly after the Llama 2 family of four comparably sized models in the summer of 2023. An analysis of cumulative model downloads shows the Llama 2 models being downloaded about 500% of that of early Qwen models (a difference of 10M versus 60M total downloads with half of the models), highlighting the original state of play in the open ecosystem – a large lead for American models.
Llama 3 continued this trend with a series of models across 2024. Pieces of the Llama 3 family (and its various versions in Llama 3.1 and 3.2) are some of the most popular models ever in HuggingFace’s history as the leading distributor of open models. At the same time, the newer Qwen models from Alibaba, this time the Qwen 2.5 suite of 2024, showed substantially closer adoption numbers to Meta’s Llamas – a lead of only 20 million cumulative downloads for Llama 3 over the Qwen 2.5 suite with both of them crossing over 120M total downloads.
Llama’s lead was built on a combination of strong performance and existing distribution channels. This success came in spite of a restrictive license – the contract between the open artifact’s creator and the downstream user – that can require nuanced legal consideration about if a particular use-case is compliant. Meanwhile, Qwen and other Chinese models have adopted simpler licenses drawing on historical practices in open-source software (OSS), removing another barrier to uptake on their models.
Meta has effectively been a singular horse in this race. As language models were established as a core technology, competition has arrived. Between the last releases of Llama 3 and the arrival of Llama 4, the landscape of open models changed substantially with the arrival of DeepSeek’s permissively licensed, frontier models in DeepSeek V3 and DeepSeek R1. Now, Meta was effectively alone in releasing its best models regularly and expected to compete with Qwen making large families of models great at any size scale and DeepSeek releasing open frontier models. Both types of models are crucial to the health of the ecosystem, but they can take slightly different foci to get right.
China today has 5 amazing open labs, a number which is growing, and America has Meta as its open models champion. We are running Meta in a race against 5 other Chinese runners, and then complain when it doesn't win every race every time. Our problem is not Llama 4 being not state-of-the-art; our problem is running a solo athlete against a team built with an ecosystem to support its growth.
Chinese open models are taking the all-time lead in adoption
The available data showcasing adoption of open language models – how much models are downloaded and how much base models are modified for new uses – shows that China has taken the lead in recent adoption and will soon take the lead in all-time adoption.
We collected historical, daily download data from 6 of the leading open model providers across the world – Meta, Google, Mistral AI, Microsoft, Alibaba Qwen, and DeepSeek AI. Grouping by locality we can see America’s early lead with Llama, Europe’s surge with Mistral’s early viral releases almost surpassing the U.S. in April of 2024, and a consistent acceleration from the Chinese providers until they’re surpassing the U.S. this summer. As of August 2025, the leading U.S. and Chinese models both have around 300M total downloads on HuggingFace with the Chinese rate of growth being notably higher. The growth rate for European models has remained lower, with their cumulative downloads reaching around 100M today.
An important benefit of open models is the ability to finetune
Finetune
The process of taking a pre-trained AI model and training it further on specific data to adapt it for particular tasks or domains. This allows customization without training from scratch.
them, a process to adapt a given model to a specific purpose. This process is at the heart of academic research and important for businesses to shape a given model to their individual needs. While there are more cumulative derivatives of American models at the moment, Chinese models are gaining momentum, especially this year.
Early in 2024, Chinese models accounted for 10-30% of the new finetuned models appearing on HuggingFace. Today, derivatives of Alibaba’s Qwen models account for more than 40% of the language models appearing on HuggingFace month over month (the overall picture is quite similar to the downloads data) – and that is just one of China’s leading open model laboratories. Meta’s share of derivatives with the Llama models has dropped from a peak of nearly 50% in the fall of 2024 down to only 15% today. With far fewer open model options appearing from the U.S. or Europe, the proportion of Chinese models in the AI ecosystem is expected to continue to rise.
4. Community Preference of Open Models
In August of 2024, Chinese open models trailed both the U.S. and the rest of the world, but they quickly overcame this to take a strong lead.
Community Elo Rankings
Monthly performance rankings, Aug 2024 - Jan 2026
US
China
Other
The ATOM Project (updated 01/2026)
source: lmarena.ai
5. Best Open Models Over Time
Linear Trend Analysis showing the performance trajectory of open models by region across ArtificialAnalysis benchmarks.
Latest AA IndexPrevious AA Index
Performance Intelligence Rankings
ArtificialAnalysis Overall Intelligence, Aug 2024 - Jan 2026
USA
China
EU
The ATOM Project (updated 01/2026)
source: artificial analysis
6. Model Adoption Trends
Percentage of new finetunes and derivatives released per month by organization that released the underlying base model. Mistral AI's and Meta's early lead has been overtaken by Alibaba's Qwen models.
Derivatives per Base Model
Nov 2023 - Feb 2026
Qwen
Meta Llama
Mistral
DeepSeek
Others
The ATOM Project (updated 03/2026)
source: huggingface
7. Who's Open Models are Used Most
China's inference lead mirrors the other adoption metrics of open models — rapid growth and control following the launch of DeepSeek. Sourced from top 10 open models per month on OpenRouter, ranked by total tokens.
Token Share by Region
% of open model inference tokens, Nov 2024 - Jan 2026
USA
China
EU
The ATOM Project (updated 01/2026)
source: openrouter.ai
Click to unfurl for more data on how the open ecosystem works and is evolving.
8. Taking on DeepSeek
Many new players have emerged after the release of DeepSeek, seeking to match them in influence. All of these new players are drastically behind Qwen in adoption numbers, but OpenAI's GPT-OSS models released in Aug. 2025 and Hugging Face's SmolLM family show meaningful starts.
New Open Model Players
Cumulative Downloads, Jul 6 2025 - Dec 28 2025
DeepSeek
Hugging Face
OpenAI
Nvidia
Z.ai
Moonshot AI (Kimi)
MiniMax
The ATOM Project (updated 01/2026)
source: huggingface
9. The Qwen Ecosystem
Qwen alone has more monthly downloads than all other major open model providers combined. Others include: Meta Llama, DeepSeek, OpenAI, Mistral AI, Nvidia, Z.ai, Moonshot AI (Kimi), and MiniMax.
December 2025 Downloads
Monthly Downloads by Organization
Qwen
Meta Llama
DeepSeek
OpenAI
Mistral
Nvidia
Z.ai
Moonshot AI (Kimi)
MiniMax
The ATOM Project (updated 12/2025)
source: huggingface
10. The Qwen 3 Effect
Just 5 Qwen3 models (25.9M downloads) surpass the combined December 2025 downloads of 6 entire competitors (22.3M). Models: Qwen3-[0.6B, 1.7B, 4B (Original), 8B, & 4B-Instruct-2507]. Organizations: OpenAI, Mistral AI, Nvidia, Z.ai, Moonshot AI, and MiniMax.
5 Small Qwen Models vs 6 Competing AI Labs
December 2025 Downloads
Qwen3 Models
Mistral
OpenAI
Nvidia
Z.ai
Moonshot AI (Kimi)
MiniMax
The ATOM Project (updated 12/2025)
source: huggingface
11. DeepSeek's Large Model Lead
DeepSeek is one of the only organizations challenging Qwen in any bracket. Qwen dominates smaller models, but hasn't clawed back DeepSeek's large model lead. At 250B+, DeepSeek captures 56% of downloads compared to Qwen's 1%.
Share of Downloads at Each Size Tier
All-time downloads by organization
DeepSeek
Qwen
The ATOM Project (updated 12/2025)
source: huggingface
12. Downloads by Model Size
The 7-9B parameter range dominates downloads, a sweet spot for research and automation tasks. Models in this range account for nearly 38% of all downloads and are the most popular size to release. Across 1,152 tracked models, we've logged over 2 billion downloads.
Share of Downloads by Parameter Count
All-time downloads across tracked models
Show Count
The ATOM Project (updated 12/2025)
source: huggingface
13. America's New Entrants
Following the release of The ATOM Project, more labs have oriented towards open models (e.g. OpenAI's GPT-OSS soon after), so we're following the downloads of these newer organizations since the Summer of 2025. In this time, OpenAI and DeepSeek have logged over 50M downloads, and Qwen over 400M, so we have a long way to go.
Cumulative Downloads
Aug 2025 - Present
Nvidia
Allen AI
IBM Granite
AI21 Labs
Arcee AI
The ATOM Project (updated 01/2026)
source: huggingface
14. DeepSeek Kickstarts Chinese AI Adoption
DeepSeek R1's January 2025 release triggered rapid AI adoption in China. Within months, China's AI user share more than doubled from 8% to over 20%, making it the biggest AI market in the world (according to Microsoft data). At 20% user share, China now has an estimated 195+ million AI users.
AI User Share by Country
Share of working-age population using AI tools (3-month avg)
United States
China
The ATOM Project (updated 01/2026)
source: arXiv:2511.02781
15. Token Share by Organization
DeepSeek dominates inference, but the market is more dynamic than downloads — new entrants like MiniMax and Xiaomi rapidly gain and lose share. Sourced from top 10 open models per month on OpenRouter. "Other" includes community models, Nous, and TNG.
Open Model Inference Market Share
% of open model inference tokens, Nov 2024 - Jan 2026
DeepSeek
Meta
Mistral
Qwen
OpenAI
MiniMax
Xiaomi
Zhipu
Moonshot
Other
The ATOM Project (updated 01/2026)
source: openrouter.ai
16. User Share by Region on HuggingFace
HuggingFace user activity by region over time. China's share grew dramatically after DeepSeek R1's release in late January 2025, jumping from ~8% to nearly 28% by mid-2025.
Weekly User Share by Region
% of HuggingFace users, Mar 2023 - Aug 2025
US
Europe
China
Unaffiliated
Other
The ATOM Project (updated 02/2026)
source: HuggingFace / FT
What the ecosystem needs
We can fix this. America has the talent, compute, and capital to lead open model development – we just need to get them to the right place.
The tone for change is well represented by the White House's recent AI Action Plan, which paints a much clearer vision for the benefits of innovation and adoption globally to far outweigh the current measured risks. This represents an inflection point in the perception of open models, especially in the United States, but we still have a long way to go to support this vision with artifacts and actions.
The United States has a thriving AI research community, but it is missing the models that it itself has created and has complete knowledge of in order to create clear, and rapid progress. For example, the area of research with the most excitement following recent reasoning models is reinforcement learning with verifiable rewards (RLVR)
RLVR
A training method where AI models learn through trial and error with rewards that can be mathematically verified as correct. This is particularly useful for tasks like math and coding where answers can be checked for accuracy.
. This research has largely been performed on Alibaba's Qwen models from China due to their strong performance across math, code, and STEM benchmarks.
There are two categories of truly open models that we need in order to lead on all metrics of open models defined by how AI is studied and used. Both are essential and complement each other and the rest of a leading AI ecosystem. The best outcome is when these are accompanied by training data, intermediate checkpoints, base models, training code, and permissive licenses accepted as standards for free use by the AI community. These models with everything released, currently less common across the industry, are known as “open source models” to clearly note the benefits that come with more knowledge of how it was built.
First, we need leading open models at the frontier of performance. These should be the best models in the world and can be complementary to offerings from the leading closed AI models built in America, offering cheaper costs and more modifiability. The fundamental insight driving the recent rapid buildout of AI training infrastructure is the idea of scaling laws – this applies to open and closed models alike. The ballpark of scale needed to reach the leading edge of performance today is 100 to 600+ billion parameters with a mixture of experts (MoE) architecture – a size range used for all the leading open models from the U.S. and China in 2025 that challenge the best closed models on intelligence benchmarks.
With these leading models, we need a family of related models across a variety of sizes to allow every application and direction of study to be addressed. This is a standard adapted by leading open model suites from the U.S. and China alike. Only the most challenging tasks need the largest models, and for the rest of the tasks facing AI there needs to be tools to understand the minimum model size to solve certain simple tasks. A distribution of model sizes from those that can run on your iPhone to those that are assisting with the hardest intellectual work and everything in between creates maximum opportunity to advance and integrate AI broadly.
We estimate the entry point to train models of this size distribution is a cluster of compute on the order of 10,000+ leading GPUs. It is standard for top models to be trained with small teams of fifty to a few hundred people. A famous number on the cost of training frontier AI models from earlier this year was the often quoted $5 million figure for DeepSeek V3 – this is misleading on what it actually takes to develop these models, and the authors of the DeepSeek technical report acknowledged so much. 10,000 GPUs provide an entry point for rapid iteration concurrent to large-scale training.
America should target having multiple centers producing excellent open models. This serves to de-risk progress on training these models, given the urgency of the mission, but will also allow for a more diverse set of artifacts and for the research groups to learn from each other without first making the training organizations so large that progress is slowed.
There are many avenues to obtain and allocate these resources across multiple stakeholders. We need to engage across private companies, philanthropic institutions, and government agencies. Programs such as the National AI Research Resource (NAIRR) are important for broadening access to resources related to AI research including compute, data, software, and models, but these ecosystem-wide solutions are not enough to create breakthrough models as China is with concentrated bets. We need immediate, targeted interventions that can deliver frontier open models within 6-12 months, not years.
As many organizations around the world create strong AI models, it is becoming clearer that with the right compute and talent, strong models can follow. The formula we must follow is delivering these resources with the directive to release the models openly, then we can solidify American AI leadership. Every stakeholder – from tech giants to philanthropies to federal agencies to researchers and engineers – must ask themselves: Are we funding or participating in the future of AI research, or are we ceding it to competitors who understand that open models are the foundation of AI supremacy?
Sign on to support:
Building American Truly Open Models will take time, resources, and the support of hundreds of the finest minds in America. If you want to support us, sign up to join the cause.
Learn How You Can Help
View Signatories (350+ signed)
Frequently Asked Questions
Open ModelsAI EcosystemMiscellaneous
Open Models
Why do companies release open models?
How did China get ahead in open AI models despite chip restrictions?
What's the difference between open and closed AI models?
What about open vs. open weight vs. open source models?