How Many AI Tools Are There? 51,242 Apps, 2.9M Models, One Honest Answer

There is no single number, because “AI tool” is not one thing. As of July 17, 2026, curated directories list roughly 51,242 consumer AI tools (There’s An AI For That), Hugging Face hosts about 2.9 million downloadable models, GitHub carries around 219,362 repositories under its largest AI topic, and PyPI lists 10,000+ AI Python packages. These pools overlap heavily and can never be added into one total. Ask “how many AI tools are there” and every answer between “a few thousand” and “several million” is technically correct depending on which layer of the stack you count. The opponent this post argues against is every article that hands you one confident total.

Quick reference: the short answer by layer

#	What is being counted	Source	As of	Count
1	Consumer AI tools	There’s An AI For That	Jul 17, 2026	51,242
2	AI models hosted	Hugging Face	Jul 17, 2026	2.9M+
3	Repos tagged “machine-learning”	GitHub topics	Jul 17, 2026	219,362+
4	AI Python packages	PyPI	Jul 17, 2026	10K+
5	AI datasets	Hugging Face	Jul 17, 2026	962,657
6	Hosted AI demos (Spaces)	Hugging Face	Jul 17, 2026	1,422,199
-	Futurepedia catalog	Futurepedia	Jul 17, 2026	4,000+
-	OpenTools catalog	OpenTools	Jul 17, 2026	2,500+

What actually counts as an “AI tool”

The count changes wildly depending on the layer you measure. A packaged app a person clicks is one definition. A set of trained weights a developer downloads is another. Six layers that people mix up when they quote a number:

Term	What it is	Example	Rough count
Consumer AI tool	Packaged app or service an end user interacts with directly	ChatGPT, Midjourney, Perplexity	~51K
AI model	Set of trained weights you can download and run yourself	Llama 3, Stable Diffusion, Whisper	~2.9M
Open-source AI project	Public code repository tagged with an AI or ML topic	PyTorch, LangChain, Ollama	~219K
AI Python package	Installable library classified under the AI topic on PyPI	transformers, scikit-learn, langchain	10K+
AI dataset	Training or evaluation data collection hosted for reuse	ImageNet, The Pile, Common Crawl	~963K
AI demo / Space	Hosted, runnable ML app or demo	Hugging Face Spaces	~1.4M

Stack the ecosystem by layer and the spread is enormous. A few tens of thousands of polished consumer apps sit on top of millions of raw models and projects. Most of what people call a “new AI tool” is a thin wrapper around a model in the layer beneath it.

Hugging Face is the biggest countable layer

Hugging Face is the largest open hub for hosted models, datasets, and Spaces. It dominates the machine-countable AI pool, dwarfing every consumer directory.

Metric	Value
Models	2,918,668
Datasets	962,657
Spaces (hosted demo apps)	1,422,199

GitHub topics: overlapping counts by definition

GitHub tags repositories by topic. A single repo can carry several AI topics at once, so these counts overlap and must never be summed into one total. The “machine-learning” tag alone is larger than the entire consumer AI-tool market.

Topic	Repos
machine-learning	219,362
deep-learning	99,603
llm	97,780
artificial-intelligence	41,072
generative-ai	15,954

How fast the count is growing

Both consumer tools and hosted models have climbed steeply since 2021.

Layer	2021	2024	2026	Multiplier
Hugging Face models	~30,000	~1,200,000	~2.9M	~97x
Consumer directory tools	~1,000	~15,000	51,242	~51x

The model layer is compounding roughly 1.9 times faster than the consumer-tool layer. That gap is the real story. Most of what gets called a “new AI tool” is a wrapper around a model that already existed underneath. The wrapper economy has grown 51x. The model layer that everything depends on has grown 97x.

The daily pace

There’s An AI For That added 44 new tools on July 17, 2026 alone. Sustained for a year, that pace would add roughly 16,060 tools. It swings day to day, so treat the yearly projection as a rough ceiling, not a forecast. The counter is live and represents that day’s activity, not a stable long-run average.

Five facts worth quoting

Hugging Face out-models every AI tool directory combined by 51x. Add up every tool tracked by There’s An AI For That, Futurepedia, and OpenTools and you get about 57,742 consumer AI tools. Hugging Face lists 2.9M+ trained models on its own.

There are 19x more AI datasets than consumer AI tools. There’s An AI For That lists 51,242 consumer-facing AI tools. Hugging Face hosts 962,657 datasets used to train and evaluate models.

GitHub’s “machine-learning” topic beats the consumer AI-tool market by 4.3x. GitHub tags 219,362 repositories under that single topic. That is one topic tag out of several tracked here.

The model layer is growing 1.9x faster than the tool layer. Since 2021, Hugging Face’s model count has grown ~97x while directory-tracked consumer tools grew ~51x over the same span.

44 new tools per day. The live counter on July 17, 2026. Not a stable long-run average, but a real signal of pace.

Why the “how many AI tools” number keeps moving

Every “how many AI tools exist” figure is a snapshot of one definition, taken on one day. There’s An AI For That counted 51,242 consumer-facing tools as of July 17, 2026, a number already out of date by the time you read this, because directories like it add new listings daily and quietly retire dead ones without announcing it.

The growth curve behind that number is not gentle. Hugging Face’s hosted model count went from 30,000 in 2021 to 2.9M today. Directory-tracked consumer tools grew from 1,000 in 2021 to 51,242. Real growth that tracks the broader AI adoption statistics, but slower than the model layer underneath.

GitHub tells a similar story from the builder’s side. Its “machine-learning” topic alone tags 219,362 repositories, and a repo can carry several AI topics at once, so the counts overlap and are never summed into one total. That is the honest reason nobody can hand you one clean “total AI projects” number. The same repo, the same underlying model, and the same wrapped consumer app can each get counted in a different tier on this page.

Watch the trend, not any single day’s counter, and always check what a source is actually counting before you quote its total. The same care goes into counting how many websites exist and any other “how many X” number.

Directory versus directory

Curated directories disagree because each has a different bar for inclusion. The largest counts every reviewed app. Smaller editorial directories keep a tighter, hand-picked catalog.

Directory	Listed tools	What it counts
There’s An AI For That	51,242	Every individually reviewed consumer AI tool and app
Futurepedia	4,000+	Editorial, hand-curated catalog of notable tools
OpenTools	2,500+	AI tools plus MCP servers (publisher-stated size)

For the vetted picks across categories, the best AI tools list applies the same buyer-side discipline to a much shorter list. The AI glossary covers the vocabulary underneath each layer above. For how new tool discovery happens now that AI-search citations matter more than blue links, how AI search engines work explains the retrieval mechanics.

Sources

Every number on this page carries a public source and an as-of date. Nothing here is an unsourced estimate.

Source	Measures	As of	Access
There’s An AI For That	Curated consumer AI tools	Jul 17, 2026	Public counter
Hugging Face	Hosted models, datasets, Spaces	Jul 17, 2026	Listing headers
GitHub topics	Repositories per AI topic	Jul 17, 2026	Free Search API
PyPI	AI-classified Python packages	Jul 17, 2026	Public index search
Futurepedia	Editorial AI-tools directory	Jul 17, 2026	Publisher-stated
OpenTools	AI tools and MCP directory	Jul 17, 2026	Site meta

Cite this page: “As of July 2026, curated directories list roughly 51,000 consumer AI tools, while Hugging Face hosts about 2.9 million downloadable models. There is no single count, the number depends on which layer of the AI stack you measure.” Source: zplatform.ai, How Many AI Tools Are There.