Keras VS PyTorch
As of 2 November 2026, PyTorch is named by 4,159 companies and Keras by 464; Keras skews to companies under 200 people (50% vs 57%), PyTorch to companies over 1,000 (23% vs 31%).
As of 2 November 2026 · refreshed daily · counts move as websites, DNS and career sites are re-read.
The website-and-DNS footprint is not comparable between a marketing suite and a CRM: Keras's script and mail records show on hundreds of thousands of small sites, PyTorch rarely leaves a trace outside its own apps. Compare on the job-posting numbers; use the footprint for reach.
- Keras: 2–200 employees
- 50%
- PyTorch: 2–200 employees
- 57%
- Keras: 1,000+ employees
- 31%
- PyTorch: 1,000+ employees
- 23%
- Postings per company, 90 days
- 2.8 · 4.2
- Website or DNS footprint
- 0 · 0
Who chooses which
Share of each tool's companies in every segment. "Leans" marks a gap of more than two points.
| Segment | Keras | share | PyTorch | share | Leans |
|---|---|---|---|---|---|
| Self-employed | 6 | 1% | 112 | 3% | even |
| 2-10 employees | 37 | 8% | 589 | 14% | PyTorch |
| 11-50 employees | 110 | 24% | 928 | 22% | even |
| 51-200 employees | 87 | 19% | 839 | 20% | even |
| 201-500 employees | 52 | 11% | 454 | 11% | even |
| 501-1000 employees | 26 | 6% | 293 | 7% | even |
| 1001-5000 employees | 52 | 11% | 455 | 11% | even |
| 5001-10000 employees | 19 | 4% | 127 | 3% | even |
| 10,001+ employees | 75 | 16% | 362 | 9% | Keras |
| Segment | Keras | share | PyTorch | share | Leans |
|---|---|---|---|---|---|
| Technology & software | 238 | 51% | 2,236 | 54% | PyTorch |
| Business & finance services | 69 | 15% | 618 | 15% | even |
| Manufacturing & industry | 69 | 15% | 552 | 13% | even |
| Consumer services & leisure | 22 | 5% | 126 | 3% | even |
| Media & content | 17 | 4% | 131 | 3% | even |
| Public, community & associations | 15 | 3% | 171 | 4% | even |
| Health care | 10 | 2% | 54 | 1% | even |
| Commerce & trade | 7 | 2% | 66 | 2% | even |
| Utilities & infrastructure | 6 | 1% | 75 | 2% | even |
| Transport, logistics & rental | 4 | 1% | 35 | 1% | even |
| Automotive | 3 | 1% | 41 | 1% | even |
| Property & construction | 3 | 1% | 33 | 1% | even |
| Segment | Keras | share | PyTorch | share | Leans |
|---|---|---|---|---|---|
| United States | 284 | 61% | 2,568 | 62% | even |
| United Kingdom | 27 | 6% | 274 | 7% | even |
| India | 26 | 6% | 201 | 5% | even |
| Germany | 17 | 4% | 212 | 5% | even |
| Canada | 12 | 3% | 96 | 2% | even |
| France | 8 | 2% | 106 | 3% | even |
| Poland | 7 | 2% | – | – | outside top 12 |
| Singapore | 7 | 2% | 42 | 1% | even |
| Australia | 6 | 1% | 49 | 1% | even |
| Spain | 6 | 1% | 48 | 1% | even |
| Italy | 6 | 1% | 40 | 1% | even |
| Netherlands | 6 | 1% | 51 | 1% | even |
| Switzerland | – | – | 46 | 1% | outside top 12 |
| Segment | Keras | share | PyTorch | share | Leans |
|---|---|---|---|---|---|
| Data & AI | 1,328 | 71% | 16,026 | 69% | Keras |
| Software engineering | 124 | 7% | 2,282 | 10% | PyTorch |
| Unclassified | 47 | 3% | 147 | 1% | even |
| Research and science | 24 | 1% | 475 | 2% | even |
| Engineering (non-software) | 20 | 1% | 515 | 2% | even |
| Design & UX | 19 | 1% | 35 | 0% | even |
| Legal, compliance & risk | 15 | 1% | 25 | 0% | even |
| General management & consulting | 14 | 1% | 94 | 0% | even |
| Education & training | 13 | 1% | 55 | 0% | even |
| Product, program & project management | 9 | 0% | 162 | 1% | even |
| IT operations & business systems | 8 | 0% | 444 | 2% | even |
| Sales | 8 | 0% | 211 | 1% | even |
| Arts media and entertainment | 6 | 0% | 10 | 0% | even |
| Administration & office | 5 | 0% | 7 | 0% | even |
| Customer service & success | 5 | 0% | 37 | 0% | even |
| Human resources | 5 | 0% | 22 | 0% | even |
| Finance & accounting | 4 | 0% | 67 | 0% | even |
| Healthcare clinical | 4 | 0% | 5 | 0% | even |
| Manufacturing and production | 2 | 0% | 4 | 0% | even |
| Marketing & communications | 2 | 0% | 39 | 0% | even |
| Agriculture and environment | 1 | 0% | 5 | 0% | even |
| Construction and civil | 1 | 0% | 1 | 0% | even |
| Skilled trades and field service | 1 | 0% | 14 | 0% | even |
| Business development & strategy | – | – | 29 | 0% | outside top 12 |
| Quality, safety & environment | – | – | 12 | 0% | outside top 12 |
| Logistics warehouse and transport | – | – | 7 | 0% | outside top 12 |
| Hospitality and food service | – | – | 2 | 0% | outside top 12 |
| Real estate and property | – | – | 2 | 0% | outside top 12 |
| Facilities cleaning and security | – | – | 1 | 0% | outside top 12 |
| Procurement and purchasing | – | – | 1 | 0% | outside top 12 |
The companies behind each column
Companies using Keras →
Companies using PyTorch →
ml frameworks ranked →
Quote the sentence with the date and the link. The numbers change daily; the link always shows the current ones.
PyTorch is named by 4,159 companies in job postings and Keras by 464; 746 name both. (Breaknorm, 2 November 2026, https://breaknorm.co/compare/keras-vs-pytorch)
Read from the source, not self-reported
Every number on this page can be traced to a page we crawled or a posting we read. Full methodology →
- Job postingsRead hourly from career sites. A tool counts when the employer's own text names it as the team's system; both tools can count for one company.
- Website and DNSScripts and embeds on the company's pages, plus SPF, DKIM, MX and CNAME records of its domain that point at the vendor.
- SegmentsCountry, size and industry from the company profile built from its website.
- Not a reviewNo ratings, prices or feature claims. For those, read the vendors.
Keras vs PyTorch: questions
Which is more used, Keras or PyTorch?
By companies naming it in their own job postings, PyTorch (4,159) over Keras (464). By website-and-DNS footprint, PyTorch (0 vs 0), because its scripts and mail records sit on many small company sites.
Who chooses Keras?
50% of Keras companies have 2 to 200 employees, against 57% for PyTorch. 31% of Keras companies have over 1,000 employees, against 23% for PyTorch.
Do companies run both?
746 companies name both in their postings, usually Keras in marketing and PyTorch in sales, or during a migration.
Where do these numbers come from?
From the employer's own text in job postings read hourly from career sites, and from scripts found on company websites in our crawl. No vendor data, no surveys.
Get either list your way
The same data as a search, over MCP and the API, as bulk files, or pushed to you as it changes.
Every company on either tool,
with the person who owns it
Filter on Keras or PyTorch. Start free, no card.