brushllm.com is live
Welcome to BrushLLM — one OpenAI-compatible endpoint for frontier models from ten Chinese AI labs. Grab an API key and start building in under a minute.
Read moreNew Kimi K2 Thinking & Qwen3-Max are live. Browse models →
24 models · 10 Chinese labs · one endpoint
BrushLLM connects developers in every country directly to China’s frontier models — DeepSeek, Qwen, GLM, Kimi, Doubao and more — through a single OpenAI-compatible endpoint.
Free to start · Pay-as-you-go · Set up in 30 seconds
Why BrushLLM
A straight path from “somewhere outside China” to “calling DeepSeek in production”.
Keep your existing SDK and code. Change one base URL and every request just works — chat, tools, and streaming.
DeepSeek, Qwen, GLM, Kimi, Doubao, MiniMax, ERNIE, Hunyuan, Spark and StepFun — 24 models behind a single API key.
No mainland-China accounts, local phone numbers, or separate billing per lab. Sign up from anywhere on earth and start calling in under a minute.
Per-token billing with no subscriptions and no minimums. Several models — including GLM-4.5 Flash and Spark Lite — are free.
Create multiple API keys, set spend quotas per key, and monitor usage from the console — safe to hand keys to agents and teammates.
Upstream labs are health-checked around the clock, with automatic failover so a single provider outage never takes your app down.
How it works
Sign up with email and grab an API key from the console. No credit card to start.
Set the base URL and keep the OpenAI SDK you already use.
Swap model ids freely — the whole catalog is one key away.
base_url = https://api.brushllm.com/v1
Drop-in compatible
If it works with the OpenAI API, it works with BrushLLM. Here’s the same request in four languages.
curl https://api.brushllm.com/v1/chat/completions \
-H "Authorization: Bearer $BRUSH_LLM_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-chat",
"messages": [
{"role": "user", "content": "Hello, BrushLLM!"}
]
}'
10+
Chinese AI labs
24+
Models, one catalog
1
API key
195
Countries served
AI News
New releases, catalog updates and platform news — refreshed daily. Click any story for details.
Welcome to BrushLLM — one OpenAI-compatible endpoint for frontier models from ten Chinese AI labs. Grab an API key and start building in under a minute.
Read moreDeepSeek-V3.2, Qwen3-Max, GLM-4.6, Kimi K2 and more are all available today, including three free tiers for prototyping.
Read moreBuild search and RAG pipelines with text-embedding-v4 through the same base URL and API key as chat models.
Read moreYou can now cap monthly spend per key from the console — handy when you hand keys out to side projects or teammates.
Read moreIndicative per-token prices for every model in the catalog are now listed in one table. Log in to the console for your actual rates.
Read moreThe docs now cover authentication, SDK examples in four languages, and a troubleshooting table for common errors.
Read moreFAQ
BrushLLM is the source of China’s frontier AI for the world: a gateway that connects developers in every country directly to the leading Chinese labs — DeepSeek, Qwen, GLM, Kimi and more — through one OpenAI-compatible endpoint.
Yes. Point your SDK’s base URL at https://api.brushllm.com/v1 and use your BrushLLM key. Chat completions, tool calling, and streaming work the way you expect — only the model ids change.
Strictly pay-as-you-go per token, in USD. There are no subscriptions or minimums, and several models have free tiers. Top up and track spend from the console.
No. BrushLLM is built for developers everywhere: you sign up with an email, pay in USD, and reach every model through our endpoint — no local accounts or special network setup required.
DeepSeek-V3.2 is a great default for chat and tools, GLM-4.6 and Qwen3-Max excel at coding and agentic workflows, and DeepSeek-R1 or Kimi K2 Thinking for hard reasoning. Browse the full catalog with live prices on the Models page.
Yes — anything that speaks the OpenAI API works. Set the base URL and API key in your tool’s config (Cursor, Cline, OpenCode, aider and similar) and pick any model id from the catalog.
One key, 24 models, 10 labs — from the source to your code, anywhere in the world.
Get my API key