// Chapter 01 · Ecosystem
Each Source Has a Different Leader
Qwen takes more than half of the downloads we track on Hugging Face, mostly through small models. We assume they run on home GPUs. But is it what companies use?
Data: Hugging Face download and repository statistics (cfahlgren1/hub-stats, Apache-2.0), ATOM Project tracked-model list, OpenAlex (CC0).
54%
of the downloads of the models we track on Hugging Face from October 2025 to September 2026 went to Qwen.
Hugging Face downloads (hub-stats); ATOM Project.

From October 2025 to September 2026 Qwen accounted for 54% of the downloads of the open models we track on Hugging Face, and eight of the ten most downloaded repositories are its own. Most of those downloads are of small models, which may be easier to distribute and to try on home hardware. But a model that is downloaded a lot is not necessarily the one used most: in the tokens served through OpenRouter, DeepSeek comes first.
01 Most downloaded
Qwen has eight of the year’s ten most downloaded models
The ranking is led by small models. The most downloaded, Qwen3-0.6B, has 0.6 billion parameters; twelve of the fifteen have fewer than 10 billion.
The most downloaded open models in our selection on Hugging Face
Qwen Other developers
October 2025 to September 2026 · share of all downloads of the models we track
- 1
Qwen3-0.6BQwen201.9 million 5.1%
- 2
Qwen2.5-7B-InstructQwen137.5 million 3.5%
- 3
Qwen3-8BQwen107.5 million 2.7%
- 4
Qwen2.5-1.5B-InstructQwen103.5 million 2.6%
- 5
Llama-3.1-8B-InstructMeta96.6 million 2.4%
- 6
Qwen2.5-3B-InstructQwen92.9 million 2.3%
- 7
gpt-oss-20bOpenAI81.1 million 2.0%
- 8
Qwen2.5-VL-3B-InstructQwen72.6 million 1.8%
- 9
Qwen3-4BQwen70.6 million 1.8%
- 10
Qwen2.5-VL-7B-InstructQwen68.8 million 1.7%
Show places 11 to 15
- 11
Llama-3.2-1B-InstructMeta68.0 million 1.7%
- 12
Qwen3-VL-8B-InstructQwen67.0 million 1.7%
- 13
Qwen3.5-9BQwen62.7 million 1.6%
- 14
gemma-4-26B-A4B-itGoogle62.0 million 1.6%
- 15
gemma-4-31B-itGoogle58.7 million 1.5%
July to September 2026 · share of all downloads of the models we track
- 1
Qwen3-0.6BQwen80.4 million 6.1%
- 2
Qwen3-8BQwen41.1 million 3.1%
- 3
Qwen3.5-9BQwen33.9 million 2.6%
- 4
gemma-4-26B-A4B-itGoogle33.4 million 2.5%
- 5
Qwen2.5-7B-InstructQwen32.4 million 2.5%
- 6
Qwen3-VL-8B-InstructQwen30.6 million 2.3%
- 7
gemma-4-31B-itGoogle30.4 million 2.3%
- 8
Qwen2.5-1.5B-InstructQwen29.4 million 2.2%
- 9
Qwen3.6-35B-A3B-FP8Qwen28.0 million 2.1%
- 10
Llama-3.2-1B-InstructMeta24.4 million 1.8%
Show places 11 to 15
- 11
Qwen2.5-VL-7B-InstructQwen23.5 million 1.8%
- 12
gpt-oss-20bOpenAI21.5 million 1.6%
- 13
Qwen3.5-4BQwen21.2 million 1.6%
- 14
Llama-3.1-8B-InstructMeta20.0 million 1.5%
- 15
Qwen2.5-0.5B-InstructQwen20.0 million 1.5%
September 2026 · share of all downloads of the models we track
- 1
Qwen3-0.6BQwen29.6 million 7.3%
- 2
Qwen3-VL-8B-InstructQwen16.6 million 4.1%
- 3
gemma-4-26B-A4B-itGoogle12.8 million 3.2%
- 4
Qwen3-8BQwen11.7 million 2.9%
- 5
gemma-4-31B-itGoogle9.9 million 2.4%
- 6
Qwen2.5-7B-InstructQwen9.7 million 2.4%
- 7
Qwen3.5-9BQwen9.5 million 2.3%
- 8
Qwen2.5-0.5B-InstructQwen8.3 million 2.1%
- 9
Qwen2.5-1.5B-InstructQwen7.7 million 1.9%
- 10
Llama-3.2-1B-InstructMeta7.6 million 1.9%
Show places 11 to 15
- 11
Qwen3.5-4BQwen7.4 million 1.8%
- 12
Qwen3.8-27BQwen7.0 million 1.7%
- 13
gpt-oss-20bOpenAI6.8 million 1.7%
- 14
Qwen3.6-35B-A3B-FP8Qwen6.6 million 1.6%
- 15
Llama-3.1-8B-InstructMeta6.2 million 1.5%
Without Qwen: who leads the downloads
Leaving Qwen out: the most downloaded families and models, last 12 months
- 1
Gemma 4 Google
201.7 million 5.1%
- 2
gpt-oss OpenAI
131.1 million 3.3%
- 3
Llama 3.2 Meta
128.1 million 3.2%
- 4
Llama 3.1 Meta
124.4 million 3.1%
- 5
DeepSeek-R1 DeepSeek
91.3 million 2.3%
- 6
Gemma 3 Google
90.1 million 2.3%
- 7
Mistral 7B Mistral AI
70.0 million 1.8%
- 8
Nemotron 3 NVIDIA
63.4 million 1.6%
- 1
Llama-3.1-8B-InstructMeta96.6 million 2.4%
- 2
gpt-oss-20bOpenAI81.1 million 2.0%
- 3
Llama-3.2-1B-InstructMeta68.0 million 1.7%
- 4
gemma-4-26B-A4B-itGoogle62.0 million 1.6%
- 5
gemma-4-31B-itGoogle58.7 million 1.5%
- 6
gpt-oss-120bOpenAI50.0 million 1.3%
- 7
DeepSeek-OCRDeepSeek34.3 million 0.9%
- 8
gemma-3-1b-itGoogle31.9 million 0.8%
Without Qwen, the most downloaded families of the year come from US labs: Gemma 4 (Google), gpt-oss (OpenAI) and Llama 3.2 (Meta). None comes close: Gemma 4, the largest, took 5% of all the downloads, about a tenth of Qwen’s 54%, and Llama 3.2 and Llama 3.1 together 6%.
02 Downloads
The labs behind the most downloaded models have changed
Since August 2024 the balance between labs has shifted. Almost all of the most downloaded open models now come from the United States or China. But since August 2025 models from developers based in China have taken the larger share every month. In October 2026 they took 68% of the downloads.
Share of downloads of models developed in China
68%
in October 2026, against 17% in August 2024. Each square is one percentage point.
Figure 1.1
Share of downloads of tracked open models, by developer region
From August 2025, developers based in China have led those based in the US every month.
- China-based 68%
- of which Qwen 55%
- US-based 30%
- Europe-based 2%
- China-based, without Qwen 29%
- China-based, models present in Aug 2024 23%
Source Hugging Face Hub download counters, via the cfahlgren1/hub-stats archive; ATOM Project model list (Apache-2.0)
Qwen explains much of the change, but not all of it. Leave Qwen’s models out and the other developers based in China still rise, from 9% to 29% of the remaining downloads, mostly through DeepSeek and Z.ai. Without Qwen, though, US-based labs would still lead: in October 2026 they took 66% of the downloads that did not go to Qwen.
How we measured it
Without the cap the first month starts at 24%, not 17%. Grey band, December 2024 to February 2025: counting jump (Mistral), not interpreted.
Alternative populations, not a margin of error, in the data table: without Qwen’s models, and the 486 models that existed in August 2024.
- Source
- Hugging Face Hub download counters, via the cfahlgren1/hub-stats archive (archive licensed Apache-2.0); model list and region labels: ATOM Project tracked-models @d4b1709 (Apache-2.0)
- Base
- The 1,676 tracked models (ATOM list of models with published weights) with downloads on 1 October 2026
- Evidence
- E3our tabulation
Files, checks and the full method: Evidence & Methods →
| Month | Models | China-based | of which Qwen | US-based | Europe-based | China, uncapped | China without Qwen | China, 486 models of Aug 2024 |
|---|---|---|---|---|---|---|---|---|
| 2024-08 | 441 | 17.2% | 9.5% | 68.8% | 11.8% | 23.7% | 8.5% | 17.2% |
| 2024-09 | 475 | 19.7% | 13.1% | 65.7% | 12.7% | 24.7% | 7.6% | 19.4% |
| 2024-10 | 550 | 26.1% | 15.2% | 64.4% | 8.2% | 30.7% | 12.9% | 26.4% |
| 2024-11 | 581 | 47.0% | 42.8% | 46.7% | 5.5% | 47.5% | 7.2% | 20.4% |
| 2024-12 | 632 | 49.1% | 45.1% | 46.4% | 3.8% | 55.4% | 7.2% | 25.1% |
| 2025-01 | 674 | 24.5% | 19.6% | 45.2% | 29.4% | 14.9% | 6.1% | 15.8% |
| 2025-02 | 732 | 27.0% | 18.3% | 50.8% | 21.0% | 24.1% | 10.6% | 17.9% |
| 2025-03 | 763 | 40.5% | 19.2% | 50.3% | 8.5% | 39.8% | 26.4% | 20.0% |
| 2025-04 | 803 | 41.9% | 23.0% | 48.9% | 8.4% | 41.9% | 24.5% | 22.2% |
| 2025-05 | 890 | 45.0% | 28.5% | 48.0% | 6.1% | 44.8% | 23.1% | 21.9% |
| 2025-06 | 938 | 45.7% | 32.4% | 47.9% | 5.6% | 46.9% | 19.7% | 17.0% |
| 2025-07 | 939 | 46.4% | 32.6% | 47.3% | 5.5% | 47.4% | 20.5% | 17.1% |
| 2025-08 | 1096 | 57.3% | 45.7% | 38.5% | 3.8% | 61.8% | 21.3% | 29.6% |
| 2025-09 | 1163 | 53.1% | 40.3% | 44.1% | 2.4% | 49.7% | 21.6% | 28.9% |
| 2025-10 | 1211 | 56.2% | 44.3% | 41.1% | 2.4% | 56.1% | 21.4% | 23.7% |
| 2025-11 | 1281 | 60.0% | 47.8% | 35.6% | 4.2% | 60.5% | 23.4% | 20.8% |
| 2025-12 | 1307 | 63.8% | 50.0% | 32.3% | 3.8% | 64.7% | 27.7% | 21.5% |
| 2026-01 | 1382 | 61.4% | 47.3% | 34.3% | 4.1% | 62.2% | 26.8% | 22.3% |
| 2026-02 | 1390 | 66.2% | 55.6% | 30.6% | 3.1% | 67.2% | 23.8% | 23.2% |
| 2026-03 | 1420 | 68.2% | 56.3% | 28.0% | 3.7% | 69.0% | 27.1% | 21.9% |
| 2026-04 | 1464 | 67.5% | 56.0% | 29.5% | 2.9% | 67.6% | 26.0% | 22.7% |
| 2026-05 | 1521 | 67.6% | 55.0% | 29.2% | 3.1% | 77.4% | 27.9% | 24.2% |
| 2026-06 | 1539 | 66.8% | 54.3% | 30.0% | 3.0% | 70.4% | 27.4% | 26.0% |
| 2026-07 | 1580 | 66.3% | 53.8% | 31.3% | 2.2% | 65.8% | 27.0% | 30.6% |
| 2026-08 | 1599 | 65.0% | 53.7% | 32.3% | 2.5% | 64.8% | 24.5% | 25.3% |
| 2026-09 | 1643 | 68.6% | 57.3% | 29.2% | 2.0% | 68.0% | 26.5% | 23.3% |
| 2026-10 | 1676 | 68.2% | 55.4% | 29.6% | 2.0% | 67.0% | 28.7% | 23.4% |
03 Qwen's models
Which Qwen models are downloaded most
All ten of Qwen’s most downloaded models over the year have fewer than 10 billion parameters. Three of them show the range: a tiny text model, a mid-size model tuned for instructions and one that also reads images.
Three of Qwen’s most downloaded models, last 12 months
Qwen/Qwen3-0.6BA small text model
201.9 million downloads
Qwen/Qwen2.5-7B-InstructA mid-size model tuned for instructions
137.5 million downloads
Qwen/Qwen2.5-VL-3B-InstructA model that also reads images
72.6 million downloads
Two generations carry the family: among Qwen’s models in the year’s top 100, Qwen3 (2025) took 44% of their downloads and Qwen2.5 (2024) 35%. Older models keep being fetched long after newer ones arrive.
04 Downloaded and used
The most downloaded are not the most used
From October 2025 to September 2026 Qwen took 54% of the downloads we track; in the same months it took 1.0% of the open-weight tokens in OpenRouter’s daily top 50, where DeepSeek led with 31%. Downloads and use do not always go together.
Each source has a different leader: Qwen in downloads, DeepSeek in tokens and Meta in requests
The first five labs of each source, each on its own scale. A lab has the same colour in all three columns; hover over one to see it in the others.
DownloadsHugging Face, models we trackOct. 2025 – Sept. 2026
- 1Qwen54.2% of downloads
- 2Google8.5% of downloads
- 3Meta8.2% of downloads
- 4DeepSeek6.1% of downloads
- 5OpenAI3.3% of downloads
TokensOpenRouter, open weights in the daily top 50Oct. 2025 – Sept. 2026
- 1DeepSeek31.3% of tokens
- 2Tencent16.5% of tokens
- 3Xiaomi12.2% of tokens
- 4Z.ai12.1% of tokens
- 5MiniMax7.3% of tokens
RequestsCloudflare Workers AI, text generationOct. 2026
- 1Meta29.1% of requests
- 2Z.ai17.9% of requests
- 3DeepSeek11.0% of requests
- 4Qwen8.0% of requests
- 5OpenAI7.0% of requests
Each measure crowns a different lab: Qwen leads the downloads (54%), DeepSeek the tokens (31%) and Meta the requests on Workers AI (29%). Only DeepSeek is among the four leaders of all three. There is no single answer to which open model is used most: downloads, API traffic and a platform like Cloudflare each see different uses.
05 What gets built on top
What is being built on top of Qwen
Hugging Face lets us follow the models that are published from other models. A quantised version, an adaptation for one task or a model tuned on new data can keep a reference to the original. Following those references, we found 470,289 derivative repositories in our selection. Nearly half, 229,596, start from Qwen.
Figure 1.4
Derivative repositories and publishing accounts by model family
- China-based base model
- US-based
- Europe-based
- from accounts with 1,000+ derivatives
Source Hugging Face repository metadata (hub-stats), 22 October 2026
Counted by the accounts that publish them, Qwen still leads: 35,098 accounts built on Qwen, more than on Meta’s Llama (15,496) and Google’s models (12,457) together. The shift is recent: in 2024, 8,446 accounts published derivatives of US-based models and 2,040 of China-based ones; in 2026, to 22 October, it was 13,352 against 25,574. An account is not a person, and some publish automatically.
How we measured it
- Source
- Hugging Face repository metadata (base_model declarations) via the cfahlgren1/hub-stats archive, snapshot of 2026-10-22 (archive licensed Apache-2.0); region labels: ATOM
- Base
- 470,289 derivative repositories alive on 2026-10-22
- Evidence
- E3our tabulation
Files, checks and the full method: Evidence & Methods →
| Family / year | Base region | Repositories | From accounts with 1,000+ | Share | Distinct accounts | Largest account | Its repositories |
|---|---|---|---|---|---|---|---|
| Qwen | China-based | 229,596 | 77,030 | 33.6% | 35,098 | mradermacher | 12,205 |
| US-based | 72,718 | 31,772 | 43.7% | 12,457 | Krabat | 5,049 | |
| meta-llama | US-based | 65,987 | 8,041 | 12.2% | 15,496 | mradermacher | 4,663 |
| mistralai | Europe-based | 32,857 | 6,347 | 19.3% | 8,285 | mradermacher | 4,064 |
| HuggingFaceTB | US-based | 16,111 | 454 | 2.8% | 4,436 | qualiaadmin | 746 |
| deepseek-ai | China-based | 12,269 | 2,114 | 17.2% | 4,349 | mradermacher | 972 |
| microsoft | US-based | 10,762 | 1,381 | 12.8% | 3,428 | GMorgulis | 578 |
| allenai | US-based | 3,558 | 544 | 15.3% | 486 | mradermacher | 337 |
| openai | US-based | 3,458 | 704 | 20.4% | 1,405 | dementor-research | 409 |
| nvidia | US-based | 3,105 | 802 | 25.8% | 936 | mradermacher | 378 |
| 2024 · all families | China-based | 80,928 | 58,575 | 2,040 | |||
| 2024 · all families | Europe-based | 12,358 | 1,818 | 3,433 | |||
| 2024 · all families | US-based | 64,765 | 30,088 | 8,446 | |||
| 2025 · all families | China-based | 65,438 | 10,122 | 13,615 | |||
| 2025 · all families | Europe-based | 12,766 | 3,046 | 2,704 | |||
| 2025 · all families | US-based | 60,823 | 6,732 | 14,422 | |||
| 2026 · all families | China-based | 103,252 | 11,807 | 25,574 | |||
| 2026 · all families | Europe-based | 6,316 | 1,457 | 2,044 | |||
| 2026 · all families | US-based | 55,577 | 7,905 | 13,352 |
06 Other leaders
Beyond the downloads, DeepSeek, Z.ai and Moonshot stand out for other reasons
DeepSeek was the open lab with the most tokens on OpenRouter in 19 of the 21 complete months since January 2025, and took 31.3% of the open-weight tokens from October 2025 to September 2026. Z.ai, with GLM, took 12.1%, and GLM-5.3 Flash was the most used version in the top 50 in September 2026. Moonshot barely registers in that traffic, 3.7%, but its Kimi K3.1 Pro is the open model closest to the closed ones on the tests of chapter 02.
Who stands out on each dimension
Downloads
Qwen
54.2% from Oct. 2025 to Sept. 2026; 55.4% in the October 2026 reading
Hugging Face, the models we track
Derivatives
Qwen
229,596 of 470,289 derivative repositories
Hugging Face, repositories that declare a base model
Tokens
DeepSeek
31.3% from Oct. 2025 to Sept. 2026; first in 19 of 21 months
OpenRouter, open weights in the daily top 50
Requests
Meta (Llama)
29.1% in October 2026
Cloudflare Workers AI, text generation
Capability
Moonshot (Kimi K3.1 Pro)
2.2 points from the best closed model in graduate-level science and 1.5 in long agentic tasks
Epoch AI and Vals AI, best open model
Cost per solved task
DeepSeek and Z.ai (Flash versions)
DeepSeek V4.1 Flash, $1.70 with 31.8% of tasks; GLM-5.3 Flash, $3.26 with 37.4%
Vals AI, Terminal-Bench 4.0
Research
GPT and ChatGPT; then Llama
65% and 23% in 2025; DeepSeek, from 0% to 14% since 2023
OpenAlex, papers that name a family
The three profiles differ. DeepSeek combines volume and continuity: its versions change, but the lab stays in front. Z.ai competes with cheap versions that solve a reasonable share of long tasks. Kimi K3.1 Pro comes close to the closed models in graduate-level science and agentic tasks, but it still carries little of the use, and its cost is high:$7.18 per solved task on Terminal-Bench 4.0, four times as much as DeepSeek V4.1 Flash ($1.70).
07 What papers name
Which models appear in the scientific literature
Research papers tell another story. GPT and ChatGPT are still the most named, but their share of the papers that name a model family has fallen: 65% in 2025, down from 93% in 2023. Llama rose from 10% to 23%.
Share of the research papers that name a model family, by family, 2023 and 2025
2023 2025
| Model family | 2023 | 2025 |
|---|---|---|
| GPT / ChatGPT | 93% | 65% |
| Llama | 10% | 23% |
| Gemini | 1% | 18% |
| DeepSeek | 0% | 14% |
| Claude | 2% | 12% |
| Qwen | 0% | 6% |
Qwen is rising in the literature too, from 0.3% of the papers that name a family in 2023 to 6% in 2025, but that is far from its 54% of the year’s downloads. DeepSeek was named in 14%, twice as often. Papers show what researchers study; downloads, what people run.
Chapter 01 · Closing
A cross-section of the downloads
Qwen has become the dominant name in the downloads, and much of that presence is concentrated in small models, which are easier to run on home hardware. But home use does not mean business use. The tokens served through APIs, in Chapter 03, tell a quite different story.
- Qwen’s small models
- Qwen has become the dominant name in the downloads, and much of that presence is concentrated in small models, which are easier to run on home hardware.
- Downloaded and used
- Each measure crowns a different lab: Qwen leads the downloads (54%), DeepSeek the tokens (31%) and Meta the requests on Workers AI (29%).
- Home use, business use
- Home use does not mean business use. The tokens served through APIs, in Chapter 03, tell a quite different story.