Open Source AI Model Statistics 2026: Downloads, Usage and Adoption

Open source AI models are now part of everyday software work. Developers download them, build on them and run them on their own hardware. Several groups try to count that activity, and their numbers do not always match.

This article puts the main figures side by side. It covers downloads, token use, regional shifts, costs, developer surveys and company funding. Each figure comes with its source, its date and what it counts.

One point applies to all of them. A download is a file request, not a person. A token count measures work done, not copies made.

Most open models are open-weight models. That means the trained parameters are public, but the training data and code often are not.

Key figures at a glance:

• Qwen reached 942.1 million cumulative Hugging Face downloads by March 2026. Llama reached 476.0 million.
• DeepSeek served 14.37 trillion tokens on OpenRouter between November 2024 and November 2025.
• Open models handle about 33% of active AI use but earn about 4% of AI revenue, according to Mozilla.
• 79% of developers adding AI features use open models, against 71% for closed models.
• Teams that use open models reach production less often: 51% against 63% for closed models.

Open Source AI Download Statistics by Model Family

The ATOM Report tracks about 1,500 open language models on Hugging Face. It counts a download as any HTTP request to a model file endpoint. That includes automated pulls from tools such as transformers or curl.

Alibaba’s Qwen family reached 942.1 million cumulative downloads by March 2026. Meta’s Llama family reached 476.0 million in the same month. Qwen first passed Llama in September 2025, when the totals were 325.4 million and 323.7 million.

Six months later the gap was 466.1 million. In that time Qwen added about 617 million downloads, while Llama added about 152 million.

Source: The ATOM Report, Interconnects AI, cumulative Hugging Face downloads.

DeepSeek and Mistral sit far below the top two. DeepSeek passed Mistral in January 2026, at 128.2 million downloads against 124.3 million.

Mistral led the field earlier. Its share of new derivative models peaked at 57% in December 2023, driven by Mistral 7B and Mixtral. Meta’s derivative share peaked at 44% in August 2024 and stood at 11% by February 2026.

Newer U.S. labs start from a smaller base. NVIDIA’s Nemotron recorded 30.7 million downloads between August 2025 and March 2026. The Allen Institute for AI’s OLMo recorded 14.8 million and IBM’s Granite 8.6 million.

Company announcements follow different rules. Meta said Llama passed 1 billion downloads on 18 March 2025. Google said Gemma passed 1 billion on 20 August 2026, and a Google engineer noted that the count leaves out Android and Chrome integrations.

Neither company figure has been audited by an outside group. The ATOM count for Llama covers only tracked text-generation models on Hugging Face, so the two numbers cannot be compared directly.

Qwen’s monthly volume drives its lead. In February 2026 Qwen alone produced 153.6 million downloads. The next eight open model organizations produced 71.2 million combined.

Open Source AI Download Statistics by Region

ATOM assigns each model to the United States, China or Europe based on where the releasing organization is headquartered. Total tracked downloads reached 2.04 billion by March 2026, up from 339 million a year earlier.

Source: The ATOM Report, Interconnects AI, cumulative Hugging Face downloads.

No region shrank. In March 2025 the United States led China by 80 million downloads. By March 2026 China led by 428 million, and ATOM places the crossover in late July 2025.

Survey data points the same way. Mozilla’s developer survey found open-model use at 89% in Greater China and East Asia, the highest of any region. Western Europe stood at 70%.

OpenRouter traffic shows a similar change. Mozilla’s analysis found that Chinese open-weight models grew from under 2% of weekly traffic in late 2024 to 45% by April 2026.

Hugging Face’s summer 2026 report looked at model size. In almost every month of 2026, the largest open model from a Chinese lab was bigger than any model from a U.S. lab. China’s monthly ceiling ran between 754 billion and 2.78 trillion parameters.

U.S. models stayed under 130 billion parameters in five of seven months. AMD and NVIDIA each released more than 200 new model repositories in 2026, ahead of every other organization, with LiquidAI third at around 100.

Token Usage Statistics for Open Models

Downloads count how often weights are copied. Tokens count how much work a model does. OpenRouter and a16z studied more than 100 trillion tokens of real platform traffic from November 2024 to November 2025.

DeepSeek served 14.37 trillion tokens, ahead of Qwen at 5.59 trillion and Meta Llama at 3.96 trillion. The order differs from the download ranking, where Qwen leads by a wide margin.

Source: OpenRouter and a16z, State of AI, tokens routed on OpenRouter, November 2024 to November 2025.

Open-weight models reached roughly one third of OpenRouter use by late 2025. Over the one-year window, Chinese open models averaged 13.0% of weekly token volume. Rest-of-world open models averaged 13.7%, and proprietary rest-of-world models 70%.

ATOM’s own OpenRouter series shows how fast shares can move. DeepSeek V3 and R1 together made up as much as 75.6% of open model inference tokens in June 2025. That fell to 31.1% by January 2026.

Meta’s share fell from a 37.4% peak in January 2025 to zero by August 2025. Open-model leadership on this measure changes within months.

Presenc AI uses a different mix of sources and reports a smaller open share. It puts open-weight models at about 15% of inference tokens in January 2026, up from about 1% a year earlier. In that estimate DeepSeek holds about 6%, Qwen about 5% and Llama about 3%.

Open Source vs Closed AI: Usage, Revenue and Performance

Mozilla published its first State of Open Source AI report in July 2026. It found that open models handle about 33% of active AI use but capture about 4% of global AI revenue. Closed models account for the other 67% of use and 96% of revenue.

The reason is how open models earn money. Nobody pays a per-token license fee to run open weights. Revenue goes to whoever hosts the model, which is often the buyer’s own infrastructure.

Other sources give a lower usage share. MIT Sloan, as cited by TechnologyChecker, puts closed models at roughly 80% of model usage. It also says open models reach about 90% of closed-model performance at release.

The performance gap has moved over time. Mozilla’s Chatbot Arena tracking shows the average gap falling from roughly 8.04% to as low as 0.5% by August 2024. It widened again after closed labs released new reasoning models, and stood at 3.3 percentage points in March 2026.

BenchLM uses a different scale. As of 20 October 2026, 12 of its top 50 models (24%) were open-weight, and 494 of the 972 models it tracks (51%) were open-weight.

Its top open model, Xiaomi’s MiMo-V2.6-Pro, scored 73.66 out of 100 and ranked 12th. That was 13.03 points behind the top proprietary model, Claude Opus 5.5.

BenchLM lists the Arena Elo gap between the top closed and top open model at 36.8 points on 15 October 2026. It lists 29.2 points about a year earlier. The page calls this narrowing, but the numbers as printed show a larger gap, so the trend should be read with care.

The gap depends on the task. Presenc AI’s May 2026 comparison shows near parity on knowledge and coding tests, and a wide lead for closed models on hard reasoning tests.

Source: Presenc AI, Open Source vs Closed AI Market Share 2026. Values are approximate.

Single results can break the pattern. Z.ai’s GLM-5.2 scored 81.0 on Terminal-Bench 2.1, against 83.4 for GPT-5.5. On SWE-bench Pro it scored 62.1 against 58.6 for GPT-5.5.

Mozilla’s own dataset had the best open model at 67.9 on Terminal-Bench 2.1 before GLM-5.2 arrived. The report’s data was already behind the newest release on the day it was published.

Developer Adoption and Production Use Statistics

Mozilla’s report uses a SlashData survey of 1,494 developers, fielded in May 2026. Respondents were developers adding AI features to applications.

Source: Mozilla and SlashData 2026 developer survey, fielded May 2026. The first two rows overlap.

Production is where the two groups split. Among teams working with open models, 51% reach production, against 63% for closed models. Mozilla’s report page lists the open figure as 53%, while its blog post gives 51%.

SlashData’s Alvaro Ruiz Cubero attributes the gap to missing infrastructure rather than model quality. The survey asked developers what blocks open-model projects.

Infrastructure cost led at 27%. Security and compliance followed at 26%, maintenance and updates at 24% and deployment complexity at 23%. The four answers are close in weight, so no single fix removes the problem.

Use depends on the job. Presenc AI estimates that about 94% of edge and on-device work uses an open model. The share is about 85% for cost-sensitive high-volume work and about 71% for sensitive-data work in fields such as law, finance and healthcare.

Source: Presenc AI, Open Source vs Closed AI Market Share 2026.

Survey results depend on how the question is asked. Presenc AI says about 89% of enterprises use at least one open-source model in production. The Linux Foundation, as cited by TechnologyChecker, found that 63% of organizations use open-source AI and 89% use some open source in their AI stack.

Visible web deployment is much lower. TechnologyChecker scanned more than 50 million domains in May 2026. Botpress, the most common open-source AI framework it could detect, ran on 1,558 domains, against 52,682 for the closed OpenAI API and 1,349 for Google’s Dialogflow.

Of 899 companies matched to a Botpress deployment, 64% had 10 or fewer employees and 81% had fewer than 50. About 5% had 1,000 or more. The United States held about 23% of detected deployments, followed by Spain at 6.9% and India at 6%.

TechnologyChecker notes that its method cannot see backend tools such as transformers, LangChain or Ollama. Its counts reflect the browser-facing layer only.

Hugging Face Ecosystem Statistics

Hugging Face’s spring 2026 report said the platform grew to 13 million users, more than 2 million public models and over 500,000 public datasets in 2025. Its summer report tracked January to August 2026. Public model repositories rose from 2.43 million to 2.96 million, datasets from 711,000 to 1 million and Spaces from 1.00 million to 1.44 million.

Attention is concentrated in a few models. The spring report found that about half of all models had fewer than 200 total downloads, and that the top 200 models, or 0.01% of the total, held 49.6% of all downloads. The summer report put the share of models under 200 lifetime downloads at 85.6%, and said 1.5% of repositories account for 99.2% of downloads.

The two reports count differently and cover different periods. Both show the same shape, with a small group of models taking nearly all the traffic.

The Stanford AI Index 2026 counts about 5.6 million AI-related projects on GitHub and Hugging Face. Only 206,880 of them, or 3.7%, have 10 or more stars. Hugging Face model uploads more than tripled between 2023 and 2025, reaching 332,000 in a single quarter.

Source: Hugging Face, State of Open Source on Hugging Face: Spring 2026, published 17 March 2026.

The mean grew far faster than the median. Hugging Face credits quantization and mixture-of-experts designs for pulling the top end up, while small-model use stays flat.

Small models still take most downloads. Models under 1 billion parameters account for 83% of all-time downloads, and models above 100 billion account for 1%. For downloads in 2026 alone, 3% went to models above 70 billion parameters.

Likes and downloads measure different things. The top 25 repositories by 2026 downloads and the top 25 by likes shared exactly one entry. The model all-MiniLM-L6-v2 was pulled 1.55 billion times in seven months and has 5,156 likes.

Qwen has become the base that others build on. Qwen-based models account for 151,448 derivatives on the Hub, 2.6 times Meta’s total and 4.7 times the Llama repositories alone. Google follows with 82,506.

Qwen derivatives grew by roughly 180 to 210 new repositories per day in the first seven months of 2026.

Local running follows the same pattern. Qwen records 39.6 million GGUF downloads a month, against 20.8 million for Gemma and 7.5 million for Llama. GGUF is a file format used by local runners such as llama.cpp.

The fastest growth is in the runtime layer. Repositories using the gguf library rose 464%, lerobot 194% and Apple’s mlx 148%. Transformers and peft grew 16% and diffusers 21%, while all model repositories grew 21.5%.

Agents are also a new kind of user. Claude Code made up 44.4% of agent-tagged traffic to the Hub in July 2026, down from 67.8% in April. Codex rose from 10.4% to 20.8% over the same months.

Cost, Pricing and Funding Statistics

Mozilla reports that GPT-4-class inference fell from about $20 to roughly $0.40 per million tokens over 36 months. That is a 50-fold drop, which the report ties to open-weight competition. A Linux Foundation analysis cited in the report puts closed-model inference at about six times the cost of a comparable open model.

Mozilla turns that gap into an estimate of $24.8 billion in unrealized annual savings if buyers moved eligible work to open weights. As one example, Stripe cut inference costs by 73% while handling 50 million daily API calls by moving part of its AI work to open models.

Provider price lists show the spread. Presenc AI’s May 2026 comparison lists DeepSeek V4 at $0.14 per million input tokens, against $3.00 for GPT-5.5 and Claude 4.7 Opus.

Source: Presenc AI, Open Source vs Closed AI Market Share 2026, May 2026 provider prices.

Smaller closed models sit close to open prices. GPT-5.5 mini is listed at $0.15 and Gemini 3.1 Flash at $0.075. Price alone does not separate open from closed in the lower tiers.

Companies built on open models are starting to report revenue. Mozilla’s report lists Databricks at a $5.4 billion run-rate and Mistral AI at about $400 million in annual recurring revenue, a 20-fold rise in twelve months. DeepSeek is at about $220 million and has raised $7.4 billion at a valuation above $50 billion.

Mistral announced a 1.7 billion euro Series C on 9 September 2025, at a post-money valuation of 11.7 billion euros. ASML led the round. Zhipu AI and MiniMax are both pursuing Hong Kong listings in 2026.

In May 2026, IBM and Red Hat pledged $5 billion to open-source AI tooling and infrastructure. Market.us projects the open-source AI model market will grow from $19.05 billion in 2025 to $23.08 billion in 2026, about 21%.

Licenses show where the return does not come from. Of 178 Chinese releases above 20 billion parameters in 2026, 59% use Apache 2.0 and 22% use MIT. For U.S. releases in the same size band, 29% use Apache or MIT, 41% use custom terms and 30% declare nothing.

Hugging Face notes that some of the largest recent releases, such as Kimi K3 and Qwen 3.8 2.4T, have begun adding non-commercial limits or revenue-share terms. The licensing picture is changing.

Sources