A tool for running open-source large language models locally or in the cloud on macOS/Windows/Linux, which also sells a cloud compute subscription.
v0.0.1 was released on 2023-07-08, and ten days later it caught the open-sourcing of Llama 2 (7-18). Two former Kitematic (the Docker GUI, acquired by Docker) founders wrapped llama.cpp into a "Docker-style" one-line command, `ollama run llama2`; the same day Simon Willison blogged a recommendation, and a few days later Replicate's official tutorial included it, so the first users were the readers of these developer opinion leaders. After that, every new open-source model release (Code Llama, Mistral, Llama 3, DeepSeek R1, gpt-oss) became a free traffic pulse: Ollama listed it the same day, and tutorial authors default to writing `ollama run xxx`. An official Docker image, an OpenAI-compatible API, Python/JS libraries and integrations like Firebase Genkit turned it into the default base for local LLMs (67K+ integrations); Cloud models launched in 2025-09, in 2026 it hooked into coding agents like Claude Code/Codex/OpenClaw, and it monetizes with the Pro $20/month subscription.One-line growth story, compiled by us from public sources; speculation is marked separately below.
The main engine is viral spread: direct visits are 43%, more than 80% of searches are for the name ollama, and tutorial authors and third-party tools use it by default.
Every new open-source model release is a turbo boost: as soon as Llama 2, DeepSeek R1 or gpt-oss came out, Ollama listed it the same day, and tutorials and videos followed with a wave.
By estimate, nearly half of visits are existing users coming back to find models and download new versions; about 0.8 of every 100 visits click through to the checkout page.
Model: Lenny's racecar growth framework. Engine = viral spread (the open-source tool works well, then tutorial authors and third-party tools use it by default, then new users visit the site); turbo boost = every new open-source model release; lubricant = the Cloud subscription and coding agent integrations, which keep local users inside Ollama.
Real data: 9.03M visits, -11.65% month over month, channel shares (direct 42.82%; organic search 42.69% + AI 4.06% + paid search 0.01% merged as search; referrals 5.34% + organic social 3.87% + email 1.13% + display ads 0.08% merged as links), countries (China 16.71%, US 12.49%, India 9.44%, Germany 3.84%, Brazil 3.17%), desktop 85.08%, bounce rate 37.07%, average stay 4 minutes 18 seconds, 4.78 pages per visit, top referrals (github.com 26.12%, huggingface.co 16.67%, checkout.stripe.com 10.10%, quark.cn 6.44%), social sources YouTube 56.93%, Facebook 13.82%, Reddit 10.60%, X 9.41%, and outbound link destinations stripe.com 32.78% and github.com 27.86% come from Similarweb for last month; the top search terms (ollama, ollama models, olama, ollama download, ollama cloud) come from an earlier saved Similarweb record; 73,066 visits (about 73,100) to the Stripe checkout page come from quietmoney checkout-page data, with invoice.stripe.com at 0. The old channel numbers in the profile (search 42.9%, China 19.64%, brand terms 85.7%) use an earlier basis, and only the brand-term ratio is taken here for reference.
Supporting evidence: v0.0.1 was released on GitHub in 2023, ten days later Llama 2 was open-sourced, Simon Willison wrote a blog post recommending it that day, and Replicate added it to its local-running tutorial a few days later; the official blog has a same-day listing post for almost every major model release; a spoof site, ollama.org, was discussed on HN, which shows brand search volume was already large; in 2026 it raised a $65M Series B, and the company self-reports 8.9M developers and 67,000+ integrations; in 2026 it added integrations with Claude Code, Codex and OpenClaw.
Inference: How many visits each path brings is worked out by splitting the channel visits across paths (direct 3,868,000 = existing users 2,500,000 + third-party tools 600,000 + tutorials 400,000 + blog and news 368,000; search 4,223,000 = existing users 1,800,000 + tutorials 1,400,000 + third-party tools 700,000 + blog and news 323,000; links 941,000 = tutorials 400,000 + third-party tools 210,000 + blog and news 181,000 + existing users 150,000). The ratios of 5% visiting after seeing a tutorial, 50% visiting after being asked to install by a third-party tool, and 10% visiting after seeing blog and news are guesses. Much of the traffic from China probably comes from Chinese tutorials like "deploy DeepSeek locally", which is also an inference.
| Checkout | 4–6 months ago avg | 3 months ago | 2 months ago | Last month |
|---|---|---|---|---|
| checkout.stripe.com | 103,517 | 93,743 | 94,191 | 73,066 |
| invoice.stripe.com | 0 | 530 | 1,012 | 0 |
“GitHub 183.7K; 9M+ installs a month; 1B+ model downloads”Quote from the site
New-user source: Other (Direct / referral / email / paid). Search 42.9%, direct 42.8% (brand keywords 85.7%: ollama models, ollama download, ollama cloud), referral 8.6% (quark.cn 6.44%, freedidi.com 1.93%, zhihu.com 1.68%); social sources YouTube, X, Reddit. The open-source project's own word of mouth and tutorials are the main source, and the input is classed as other.
Top channel measured by Similarweb, others estimated by rank
| Attributable new users, last month | Share |
|---|---|
| Other | 100% |
Direct + brand search 86%. Direct and brand search count as viral only when the site shows sharing mechanics; the largest share wins, under 5% means Other.
| Country | Share of visits |
|---|---|
| China | 19.64% |
| United States | 13.65% |
| India | 9.5% |
| Germany | 3.69% |
| Brazil | 3.45% |
English
YouTube, X (Twitter), Reddit
14.6K keywords · brand ~86% · organic 99.99% / paid 0.01%
| Similarweb top 5 keywords |
|---|
| ollama |
| ollama models |
| olama |
| ollama download |
| ollama cloud |
217 referring sites
| Site | Share of referrals |
|---|---|
| quark.cn | 6.44% |
| freedidi.com | 1.93% |
| zhihu.com | 1.68% |
| Category | Share of referrals |
|---|---|
| Programming and Developer Software | 38.2% |
| Computers Electronics and Technology - Other | 32.39% |
| Social Media Networks | 9.83% |
| Search Engines | 2.98% |
| 1.36% |
About 73,066 last month (Similarweb estimate), #26 of 606; the month before was 95,203. Opening checkout is not paying, so this is not revenue.
New-user source: Other (Direct / referral / email / paid). Mostly referrals, email or paid, or mostly direct with no clear source Search 42.9%, direct 42.8% (brand keywords 85.7%: ollama models, ollama download, ollama cloud), referral 8.6% (quark.cn 6.44%, freedidi.com 1.93%, zhihu.com 1.68%); social sources YouTube, X, Reddit. The open-source project's own word of mouth and tutorials are the main source, and the input is classed as other.
Free $0; Pro $20/month or $200/year; Max $100/month; Team $500/month; plus usage-based billing
Dossier confidence: high. Parts 1 and 2 come from reading the site and Similarweb data; part 3 from public reporting, founder posts, Wayback snapshots, WHOIS and similar, with speculation marked.
The number: checkout visits is Similarweb's estimate of visits from the site to a checkout payment page; opening checkout is not paying, so this is not revenue. Small sites land on fixed buckets.
Data through last month. Page updated 2026-10-31.