Live — Qwen, DeepSeek, Doubao now available
Chinese AI Models.
1/7 the Cost of GPT-4o.
Wholesale API access to Alibaba Qwen, DeepSeek, and ByteDance Doubao. OpenAI-compatible endpoint — swap your base URL and save 80%.
86%
cheaper than GPT-4o
<1s
median latency
6
models available
Drop-in replacement for OpenAI
Two lines of code. No SDK changes.
# Before (GPT-4o at $2.50/M tokens)
client = OpenAI(api_key="sk-...")
# After (Qwen 2.5 72B at $0.35/M tokens)
client = OpenAI(
api_key="cnapi-your-key-here",
base_url="https://api.cnapi.io/v1"
)
response = client.chat.completions.create(
model="qwen2.5-72b-instruct",
messages=[{"role": "user", "content": "Hello!"}]
)Available Models
All OpenAI-compatible. All significantly cheaper.
| Model | Input /1M | Output /1M | vs GPT-4o | Context |
|---|---|---|---|---|
qwen2.5-72b-instruct GPT-4o level MMLU · Best value | $0.35 | $0.80 | 7× cheaper | 131K |
qwen-long 10M context · RAG & document processing | $0.09 | $0.30 | 28× cheaper | 10M |
qwen-plus Fast · Great for high-volume apps | $0.18 | $0.60 | 14× cheaper | 131K |
qwen-turbo Fastest · Lowest cost per token | $0.06 | $0.18 | 42× cheaper | 131K |
deepseek-chat DeepSeek V3 · Best for coding | $0.40 | $1.60 | 6× cheaper | 64K |
deepseek-reasoner DeepSeek R2 · Chain-of-thought reasoning | $0.80 | $3.00 | 3× cheaper | 64K |
GPT-4o reference: $2.50/M input · $10/M output
Simple Pay-as-you-go
No subscriptions. No monthly fees. Credits never expire.
Starter
$10
≈ 28M Qwen 2.5 72B tokens
- ✓ All models included
- ✓ Credits never expire
- ✓ API key instantly
POPULAR
Builder
$50
≈ 142M Qwen 2.5 72B tokens
- ✓ All models included
- ✓ Credits never expire
- ✓ API key instantly
Scale
$200
≈ 571M Qwen 2.5 72B tokens
- ✓ All models included
- ✓ Credits never expire
- ✓ API key instantly