One API key, every AI model.
Aggregating leading models such as Qwen, DeepSeek, Kimi, and GLM behind a single OpenAI-compatible API. Flexible top-ups, pay-as-you-go per-token billing.
- Sub-100ms gateway overhead
- AES-256 token storage
- Streaming SSE pass-through
Just added
Latest Qwen models now available — call them with your existing API key.
Built for production
OpenAI-compatible
Point the OpenAI SDK's base URL at the TokenGP endpoint — existing code keeps working. cURL, Python, and Node.js all supported.
Flexible top-ups
Online top-ups (Alipay / WeChat Pay, coming soon), or contact support for a corporate transfer. One balance for every model — pay-as-you-go, no subscription.
Routing and failover
Each model can be backed by multiple upstream channels with weighted, failover, or least-cost routing. 429 and 5xx responses automatically retry to backup channels.
Make your first call in 30 seconds
Already on the OpenAI SDK? Change the base URL.
from openai import OpenAI
client = OpenAI(
api_key="sk-tokengp-...",
base_url="https://api.tokengp.com/v1"
)
response = client.chat.completions.create(
model="deepseek-v4-flash",
messages=[{"role": "user", "content": "Hello, TokenGP!"}]
)
print(response.choices[0].message.content)Per-token pricing, fully transparent
Pay for what you actually use. Streaming and non-streaming priced identically.
| Model | Context | Input / 1M | Output / 1M |
|---|---|---|---|
deepseek-v4-flashRecommended | 128K | $0.14 | $0.28 |
deepseek-v4-chat | 128K | $0.27 | $1.10 |
deepseek-v4-reasoner | 64K | $0.55 | $2.20 |
qwen3.8-max | 256K | $1.49 | $4.46 |
qwen3.7-plus | 1M | $0.35 | $1.05 |
Start integrating TokenGP
Sign up by email — roughly 30 seconds to your first API call.