# The most used AI tools and model providers

As of October 2026, the most used of the 55 AI tools and model providers runsonwhat tracks is Anthropic, found at 58,195 companies, ahead of OpenAI (31,080) and Perplexity (1,292).

The AI category moves faster than any other in the catalog. It includes model providers (OpenAI, Anthropic, Google Gemini, Mistral, Cohere), voice and media models (ElevenLabs), AI search and documentation assistants (Kapa.ai, Inkeep), and the infrastructure teams build on (Pinecone, Modal, Replicate, Braintrust, LangSmith).

Most AI usage happens server-side and leaves no public trace, so these numbers count companies that verified a domain with the vendor, embed its widget or publish it on a subprocessor list. Treat them as a floor, not a market size.

## AI tools and model providers ranked by companies using them

| # | Tool | Companies | Share of category | Last 30 days | Most popular with |
|---|---|---|---|---|---|
| 1 | [Anthropic](https://www.runsonwhat.com/reverse/anthropic) | 58,195 | 60% | +212 −15 | 1,001–5,000 employees (2.6×) |
| 2 | [OpenAI](https://www.runsonwhat.com/reverse/openai) | 31,080 | 32% | +74 −9 | 1,001–5,000 employees (2.6×) |
| 3 | [Perplexity](https://www.runsonwhat.com/reverse/perplexity) | 1,292 | 1.3% | +3 −2 | 10,000+ employees (5.2×) |
| 4 | [Chatbase](https://www.runsonwhat.com/reverse/chatbase) | 1,282 | 1.3% | – | Hospitality (2.5×) |
| 5 | [TollBit](https://www.runsonwhat.com/reverse/tollbit) | 957 | 0.98% | +1 | Newspapers (195×) |
| 6 | [ElevenLabs](https://www.runsonwhat.com/reverse/elevenlabs) | 826 | 0.84% | +7 | Broadcast Media (16×) |
| 7 | [Mistral AI](https://www.runsonwhat.com/reverse/mistral) | 726 | 0.74% | +1 | Software (3.5×) |
| 8 | [MCP Registry](https://www.runsonwhat.com/reverse/mcp-registry) | 628 | 0.64% | +203 | Engineering, Product and Design (136×) |
| 9 | [Langdock](https://www.runsonwhat.com/reverse/langdock) | 507 | 0.52% | +1 | 1,001–5,000 employees (2.9×) |
| 10 | [Voiceflow](https://www.runsonwhat.com/reverse/voiceflow) | 359 | 0.37% | – | 51–200 employees (1.6×) |
| 11 | [Azure Bot Service](https://www.runsonwhat.com/reverse/azure-bot-service) | 328 | 0.34% | – | 1,001–5,000 employees (4.9×) |
| 12 | [Dust](https://www.runsonwhat.com/reverse/dust) | 199 | 0.20% | – | Software (4.2×) |
| 13 | [DeepL](https://www.runsonwhat.com/reverse/deepl) | 194 | 0.20% | +2 | 1,001–5,000 employees (8.6×) |
| 14 | [LiveKit](https://www.runsonwhat.com/reverse/livekit) | 180 | 0.18% | – | 1–10 employees (7.8×) |
| 15 | [Chatling](https://www.runsonwhat.com/reverse/chatling) | 176 | 0.18% | – | 201–1,000 employees (1.6×) |
| 16 | [Google Gemini](https://www.runsonwhat.com/reverse/gemini) | 96 | <0.1% | +1 | 1–10 employees (25×) |
| 17 | [Smithery](https://www.runsonwhat.com/reverse/smithery) | 96 | <0.1% | +3 | 1–10 employees (24×) |
| 18 | [Magnific](https://www.runsonwhat.com/reverse/magnific) | 79 | <0.1% | +2 | – |
| 19 | [Bland AI](https://www.runsonwhat.com/reverse/bland) | 68 | <0.1% | +3 −2 | – |
| 20 | [OpenRouter](https://www.runsonwhat.com/reverse/openrouter) | 55 | <0.1% | – | – |
| 21 | [Retell AI](https://www.runsonwhat.com/reverse/retell) | 49 | <0.1% | – | – |
| 22 | [Vapi](https://www.runsonwhat.com/reverse/vapi) | 45 | <0.1% | – | – |
| 23 | [Anyscale](https://www.runsonwhat.com/reverse/anyscale) | 33 | <0.1% | – | – |
| 24 | [Vizcom](https://www.runsonwhat.com/reverse/vizcom) | 33 | <0.1% | – | – |
| 25 | [Kapa.ai](https://www.runsonwhat.com/reverse/kapa) | 31 | <0.1% | −1 | – |
| 26 | [Deepgram](https://www.runsonwhat.com/reverse/deepgram) | 29 | <0.1% | – | – |
| 27 | [Modal](https://www.runsonwhat.com/reverse/modal) | 24 | <0.1% | +1 | – |
| 28 | [Unblocked](https://www.runsonwhat.com/reverse/unblocked) | 22 | <0.1% | +1 | – |
| 29 | [Groq](https://www.runsonwhat.com/reverse/groq) | 18 | <0.1% | – | – |
| 30 | [Exa](https://www.runsonwhat.com/reverse/exa) | 16 | <0.1% | – | – |
| 31 | [Firecrawl](https://www.runsonwhat.com/reverse/firecrawl) | 16 | <0.1% | – | – |
| 32 | [Langfuse](https://www.runsonwhat.com/reverse/langfuse) | 13 | <0.1% | – | – |
| 33 | [AssemblyAI](https://www.runsonwhat.com/reverse/assemblyai) | 12 | <0.1% | – | – |
| 34 | [Fireworks AI](https://www.runsonwhat.com/reverse/fireworks) | 12 | <0.1% | – | – |
| 35 | [Inkeep](https://www.runsonwhat.com/reverse/inkeep) | 12 | <0.1% | – | – |
| 36 | [LangSmith](https://www.runsonwhat.com/reverse/langchain) | 12 | <0.1% | – | – |
| 37 | [Parallel](https://www.runsonwhat.com/reverse/parallel) | 11 | <0.1% | – | – |
| 38 | [Braintrust](https://www.runsonwhat.com/reverse/braintrust) | 10 | <0.1% | – | – |
| 39 | [Cartesia](https://www.runsonwhat.com/reverse/cartesia) | 10 | <0.1% | – | – |
| 40 | [Pinecone](https://www.runsonwhat.com/reverse/pinecone) | 10 | <0.1% | – | – |
| 41 | [Cohere](https://www.runsonwhat.com/reverse/cohere) | 9 | <0.1% | – | – |
| 42 | [Clueso](https://www.runsonwhat.com/reverse/clueso) | 8 | <0.1% | −1 | – |
| 43 | [Together AI](https://www.runsonwhat.com/reverse/together) | 8 | <0.1% | – | – |
| 44 | [Baseten](https://www.runsonwhat.com/reverse/baseten) | 7 | <0.1% | – | – |
| 45 | [Glean](https://www.runsonwhat.com/reverse/glean) | 5 | <0.1% | – | – |
| 46 | [Browserbase](https://www.runsonwhat.com/reverse/browserbase) | 4 | <0.1% | – | – |
| 47 | [Helicone](https://www.runsonwhat.com/reverse/helicone) | 3 | <0.1% | – | – |
| 48 | [Replicate](https://www.runsonwhat.com/reverse/replicate) | 3 | <0.1% | – | – |
| 49 | [Tavily](https://www.runsonwhat.com/reverse/tavily) | 3 | <0.1% | – | – |
| 50 | [Weights & Biases](https://www.runsonwhat.com/reverse/wandb) | 2 | <0.1% | – | – |
| 51 | [Arize AI](https://www.runsonwhat.com/reverse/arize) | 1 | <0.1% | – | – |
| 52 | [Tavus](https://www.runsonwhat.com/reverse/tavus) | 1 | <0.1% | – | – |

Share is each tool's part of all 97,795 matches in this category; a company using two tools counts for both.

## How to compare AI tools and model providers

- Model quality on your own tasks, measured with your own evaluations.
- Data handling: retention, training on your data and regional processing.
- Price per token or per seat, and rate limits at your volume.
- Whether you need a model API, a finished assistant, or an agent that does the work.

## How runsonwhat finds AI tools and model providers

AI vendors show up in domain-verification TXT records (OpenAI, Anthropic and others ask companies to verify domains), in chat and search widgets on websites, and on published subprocessor lists.

Also tracked, no confirmed customers yet: Harvey, Hebbia, Writer.

## FAQ

### What are the most popular AI tools and model providers?

As of October 2026, the most popular AI tools and model providers by number of companies using them are Anthropic (58,195 companies), OpenAI (31,080 companies) and Perplexity (1,292 companies), according to runsonwhat. Anthropic alone accounts for 60% of the 97,795 AI matches runsonwhat has, and the top three together for 93%. 52 AI tools and model providers have at least one confirmed customer.

### Anthropic vs OpenAI: which is more popular?

Anthropic is more popular: 58,195 companies show public evidence of Anthropic, against 31,080 for OpenAI (1.9× as many), according to runsonwhat (October 2026). Perplexity is third with 1,292.

### Which AI tools and model providers do small businesses use most?

Among companies with up to 50 employees, the most used AI tools and model providers are Anthropic (18,926 companies), OpenAI (10,536 companies) and Chatbase (505 companies), according to runsonwhat (October 2026). The counts cover companies with a known headcount; see each tool's page for the full size breakdown.

### Which AI tools and model providers do large companies use most?

Among companies with more than 1,000 employees, the most used AI tools and model providers are Anthropic (5,042 companies), OpenAI (2,695 companies) and Perplexity (170 companies), according to runsonwhat (October 2026). Large companies often run more than one tool in a category, so the same company can count for several.

### Which AI tools and model providers are most popular in the United States?

In the United States, the most used AI tools and model providers are Anthropic (21,585 companies), OpenAI (11,600 companies) and Perplexity (592 companies), based on companies runsonwhat has located there (October 2026). Anthropic has 61% of the AI matches among companies based in the United States.

### Which AI tools and model providers are most popular in the United Kingdom?

In the United Kingdom, the most used AI tools and model providers are Anthropic (3,668 companies), OpenAI (1,747 companies) and Perplexity (73 companies), based on companies runsonwhat has located there (October 2026). Anthropic has 63% of the AI matches among companies based in the United Kingdom.

### Which AI tools and model providers do IT services companies use?

IT services is the industry with the most AI matches, and its companies use Anthropic (6,333 companies), OpenAI (2,858 companies) and Mistral AI (125 companies) most, according to runsonwhat (October 2026). Each tool's page breaks its customers down by industry, so you can check where a tool over-indexes.

### Which AI tools and model providers are companies adopting right now?

In the 30 days to October 2026, the most new adoptions runsonwhat recorded were for Anthropic (212 companies), MCP Registry (203 companies) and OpenAI (74 companies). A start is counted when a recheck finds a tool that the previous check of the same company, done the same way, did not.

### What is Anthropic's market share among AI tools and model providers?

Anthropic has 60% of the AI matches runsonwhat has found (58,195 companies of 97,795), as of October 2026; OpenAI has 32%. This is share of companies with public evidence of each tool, not share of revenue or seats, and a company that runs two tools counts for both.

### How does runsonwhat rank AI tools and model providers?

runsonwhat counts the companies where a public record shows each tool in use and ranks by that count. AI vendors show up in domain-verification TXT records (OpenAI, Anthropic and others ask companies to verify domains), in chat and search widgets on websites, and on published subprocessor lists. A company counts only when a direct record backs it up. The full method is at https://www.runsonwhat.com/methodology.

## Related categories

- [Frontend](https://www.runsonwhat.com/tools/frontend-frameworks.md)
- [Auth & Identity](https://www.runsonwhat.com/tools/authentication.md)
- [Backend & Database](https://www.runsonwhat.com/tools/backend-databases.md)
- [Payments & Billing](https://www.runsonwhat.com/tools/payments.md)
- [Search](https://www.runsonwhat.com/tools/site-search.md)
- [Media & UI](https://www.runsonwhat.com/tools/media-ui.md)

---

Source: runsonwhat, https://www.runsonwhat.com/tools/ai-tools (updated October 2026). Counts are companies with direct public evidence of the tool (DNS records, certificates, website code, a real-browser crawl, subprocessor lists), so they are lower bounds. Methodology: https://www.runsonwhat.com/methodology.