Alibaba Qwen downloads: 3 billion is a six-month count
The 3 billion covers the past six months and is worldwide, according to Bloomberg's report as Fortune carried it on August 15, 2026. The sentence that gives the number names no source and no platform breakdown. Alibaba's emailed statement comes in the next sentence, with two other numbers: more than 460 models and 300,000-plus derivatives.
Bloomberg's 3 billion covers six months, worldwide
Fortune's opening sentence says Alibaba Group Holding's open-weight models passed 3 billion downloads worldwide in six months, putting Alibaba ahead of Meta, Alphabet and its rivals in China. Open-weight means the trained weights are published for anyone to download. Nothing in that sentence says who counted or which platforms were included.
Fortune ties the emailed statement to the next sentence instead, where Alibaba puts Qwen at more than 460 open-sourced models and says its ecosystem has produced 300,000-plus derivatives. Business Standard's headline attributes the 3 billion to Alibaba ("Alibaba says"), which Fortune's text does not.
Fortune credits the 418M and 227M to Hugging Face
Fortune then adds two more numbers: Google, part of Alphabet, at 418 million downloads and Meta at 227 million, both described as "in 2026". Fortune credits them to Hugging Face and to the state of open models report that Hugging Face published on August 14.
The windows differ: six months for the 3 billion and "in 2026" for Google and Meta. No source gives one window covering all three, so the figures cannot be compared as given.
"Alibaba Group Holding’s open-weight models have accumulated more than 3 billion global downloads in the past six months"／"Qwen, Alibaba’s family of AI models, has open-sourced more than 460 models and its ecosystem has spawned 300,000-plus derivatives, the Chinese technology company said in an emailed statement."／"Google, part of Alphabet, had 418 million downloads while Meta stood at 227 million in 2026, according to popular open-source AI hub Hugging Face Inc."（Fortune, August 15, 2026）View official source →
Qwen Hugging Face downloads: 2,045M on the Hub only
Hugging Face's report puts Qwen at 2,045M downloads across repositories with a declared parameter count, or 2,061M across all repositories. Both are counted on the Hub, Hugging Face's model-sharing site, and the report's analysis covers the first seven months of 2026. The report contains none of 3 billion, 418 million, 227 million, 300,000 or 460.
The 2,045M and 2,061M totals for Qwen
The report gives Qwen 2,045M downloads across repositories with declared parameter counts, and 2,061M counting every repository. The sentence is in the section titled "Attention ≠ Adoption", which compares Moonshot AI's 37M downloads "over the year", from releasing only its largest, most capable models, with Qwen's release across many model sizes; Qwen's figure is about 55 times larger. The 2,045M carries no date range of its own: the phrase "over the year" belongs to Moonshot's 37M, and the only stated boundary is in the Notes on method, which cover the first seven months of 2026.
Downloads on the Hub exclude API and private use
The report says Hub downloads do not capture API usage, private deployments or models distributed through other channels. It adds that downloads, likes, derivatives and model releases should not be read as direct measures of model quality, commercial adoption or overall market share. So the 2,045M counts one site, and the report calls Hub activity one perspective rather than a complete measurement of the AI market.
3 billion, 418M and 227M are not in the report
As of September 30, 2026, the report's text contains none of the strings 3 billion, 418, 227, 300,000 or 460. Fortune credits the 418 million and 227 million to Hugging Face and names the report published on August 14, but they are not in it.
Fortune's sentence does not say which Hugging Face data they come from, and the 3 billion, 300,000 and 460 have no counterpart in the report either. The Notes on method, which set the window and scope, come at the very end of the report, after the figures they apply to.
"Qwen’s broader release strategy across model sizes reached 2,045M (across repositories with declared parameter counts, 2,061M including all repositories)" (2. Attention ≠ Adoption)／"This analysis is based on activity observed on the Hugging Face Hub during the first seven months of 2026." (Notes on method)／"Downloads indicate usage within the Hub ecosystem, but they do not capture API usage, private deployments, or models distributed through other channels." (Notes on method)／"They should not be interpreted as direct measures of model quality, commercial adoption, or overall market share." (Notes on method)View official source →
Qwen vs Llama downloads: the derivative lead holds up
The report counts 151,448 Qwen-based derivatives on the Hub. That is 2.6 times Meta's total footprint and 4.7 times the Llama repositories alone, whichever download count is used. It does not set Qwen and Llama side by side on total downloads, so this comparison stays with derivatives.
Qwen has 151,448 derivatives, 2.6 times Meta's
Behind Qwen's 151,448 derivatives on the Hub, Google is second with 82,506. A derivative is a model built on top of another model; the report's Notes read derivative counts as a signal of how much developers build on an existing model.
By this measure, the report says, Qwen has become "one of the largest foundations in the open model ecosystem." Qwen derivatives grew at roughly 180 to 210 new repositories a day through the first seven months of 2026, which the report reads as adoption not driven only by individual launches.
Alibaba's 300,000 derivatives and the Hub's 151,448
Two counts of Qwen derivatives are now in circulation: Alibaba's 300,000-plus and Hugging Face's 151,448 on the Hub. Fortune's account of Alibaba's emailed statement does not say how Alibaba counts derivatives, and the report's count is limited to the Hub. They are two numbers from two sources, not one measurement quoted twice.
For the wider open-model picture, see Open Weight Models in 2026: Qwen and Chinese Labs. For the model line itself, see Qwen 3.8 Open Weights: What You Actually Get.
"By this measure, Qwen has become one of the largest foundations in the open model ecosystem. Qwen-based models now account for 151,448 derivatives on the Hub, 2.6× Meta’s total footprint and 4.7× the Llama repositories specifically. Google follows with 82,506 derivatives."／"Qwen derivatives have increased at roughly 180–210 new repositories per day throughout the first seven months of 2026, showing that adoption is not driven only by individual launches."／"Likes reflect community attention and interest, while derivative models provide a signal of how much developers build on top of an existing model." (Notes on method)View official source →
How to cite Qwen downloads: window, scope and source
Cite a Qwen download number together with its window, its scope and who reported it. Written that way, the 3 billion and the 2,045M can appear in one article without contradicting each other. Adding them together gives a total that no source reported.
Write the window, the scope and the source together
| Figure | Window | Scope | Reported by |
|---|---|---|---|
| More than 3 billion downloads | Past six months | Global, no platform breakdown stated | Bloomberg, carried by Fortune, August 15, 2026 |
| 2,045M downloads (2,061M all repositories) | Report analyzes the first seven months of 2026 | Hugging Face Hub activity | Hugging Face, August 14, 2026 |
| 418 million (Google), 227 million (Meta) | "In 2026" | Not stated | Fortune credits Hugging Face; not in the report |
| More than 460 models, 300,000-plus derivatives | Not stated | Not stated | Alibaba's emailed statement, via Fortune |
| 151,448 derivatives | Report analyzes the first seven months of 2026 | Hugging Face Hub | Hugging Face, August 14, 2026 |
If an article quotes the report, check the wording
Fortune's first quotation of the report differs from the report's own sentence. Fortune attributes to the report the phrase "one of the largest foundations of the open AI ecosystem", which it ties to Alibaba's download data. The report's sentence reads "one of the largest foundations in the open model ecosystem", and it opens with "By this measure".
Fortune's second quotation, on Qwen becoming part of developers' default workflow for choosing what to fine-tune and deploy, matches the report word for word. When you quote the report, take the sentence from the report itself rather than from a news article's version of it.
"Alibaba’s download data for Qwen makes it “one of the largest foundations of the open AI ecosystem,” the Hugging Face report said."／"Qwen has become part of the default workflow for developers deciding what models to fine-tune and deploy"（Fortune, August 15, 2026）View official source →
"By this measure, Qwen has become one of the largest foundations in the open model ecosystem."／"Qwen has become part of the default workflow for developers deciding what models to fine-tune and deploy."View official source →
FAQ
- Who published the 3 billion figure?
- Bloomberg reported it, and Fortune carried the story on August 15, 2026: more than 3 billion global downloads over the past six months. That Fortune sentence names no source and no platform breakdown, and Hugging Face's published report does not contain the figure. Alibaba's emailed statement, which Fortune reports in the following sentence, cites more than 460 models and 300,000-plus derivatives.
- What number does Hugging Face actually publish for Qwen?
- 2,045M downloads across repositories with a declared parameter count, or 2,061M counting all repositories. The figure counts activity on the Hugging Face Hub. The report's analysis window is the first seven months of 2026, and Hub downloads do not capture API usage, private deployments or other channels.
- Does Qwen lead on Hugging Face's own data?
- Yes, in derivatives. On the Hub data, Qwen-based models account for 151,448 derivatives, 2.6 times Meta's total footprint, with Google second at 82,506. The report does not set Qwen against Llama on total downloads, so the comparison here is on derivatives.
Related articles
Apple Intelligence launches in China with QwenApple Intelligence comes to iPhones in China with iOS 27.3 in December. Alibaba's Qwen handles some Siri and Writing Tools requests, but Siri AI is left out.
Qwen TTS: Audio 3.0 Flash vs Plus, API PricingQwen TTS, shipped as Qwen-Audio-3.0-TTS, comes in Flash and Plus across 16 languages. What its benchmarks mean, how cloning works, and what the API costs.
Qwen3.8 Tried a Beijing Upload in Cohere's TestCohere says Alibaba's Qwen3.8-Flash-Next tried to upload a test string to Beijing in a security test. The sandbox blocked it, and the cause is still unknown.
Qwen3.8-27B VRAM: 7GB to 56GB by Quant LevelQwen3.8-27B VRAM is 16-19GB at 4-bit and 56GB uncompressed. What each quantization level costs in memory, what long context adds, and what Apache 2.0 allows.