LLMs
Our analysis
Most everyone uses Anthropic and OpenAI models, open-weight models from DeepSeek and Alibaba Qwen models are being used more, too.
Question
Which LLM model vendors for AI tools have you used for work in the past year? Which would you like to use next year? Select all that apply.
- scale
- Optional
- v2026.1
Data
| Respondents | Percent | |
|---|---|---|
| Anthropic | 8,871 | 74.6% |
| OpenAI | 8,834 | 74.3% |
| Google AI | 6,483 | 54.5% |
| Microsoft | 3,142 | 26.4% |
| DeepSeek | 2,253 | 18.9% |
| Alibaba Qwen | 1,203 | 10.1% |
| Mistral AI | 1,162 | 9.8% |
| xAI | 906 | 7.6% |
| Meta AI | 754 | 6.3% |
| Moonshot AI | 479 | 4.0% |
| Other | 427 | 3.6% |
| Zhipu AI | 263 | 2.2% |
| Snowflake | 234 | 2.0% |
| Databricks (DBRX) | 194 | 1.6% |
| Black Forest Labs | 120 | 1.0% |
| Cohere | 112 | 0.9% |
| Stability AI | 102 | 0.9% |
| Voyage AI (MongoDB) | 92 | 0.8% |
| Cascade | 82 | 0.7% |
| 01.AI | 56 | 0.5% |
| Luel | 41 | 0.3% |
| Guide Labs | 39 | 0.3% |
| Reka | 39 | 0.3% |
| Minimax | 13 | 0.1% |
| GLM | 4 | 0.0% |
Use this data
Response data is released under the ODbL 1.0, which asks that you attribute it.
Structured for code, with the question, the year and the source url alongside the numbers.
A formatted table, ready to paste into a document, an issue or a prompt.
Spreadsheet
.csv is a plain text format which most spreadsheet software can open.