CONTROLLED ACCESS. DEDICATED COMPUTE.
Your models.
Your perimeter.
DeepSeek, ready for your defense team. One endpoint to start compute, run inference, and understand your usage.
POWERED BY2 × NVIDIA B300DeepSeek V4.1 Flash · Abliterated 1017262k context / 64 concurrent sequences
01 / CONNECT
Enter your team API key
Your key connects you to your company’s inference and usage. Ask your administrator for access.
01 / WORKSPACE
Inference overview
COMPUTE POOLChecking
Connecting to the pool
Only start when you need it.
USAGE AT A GLANCE
—total metered tokens
—requests
—successful
—input tokens
—output tokens
Usage contains request metadata and token counts. Prompts and completions are not logged.
02 / INTEGRATE
One familiar endpoint.
Use your existing OpenAI-compatible client. Start compute before your first request; cold starts typically take about 19 minutes.
- Base URL
- https://api.grayrouter.com/v1
- Model
- deepseek-v41-abliterated
- Authentication
- Bearer YOUR_API_KEY
PYTHON
from openai import OpenAI
client = OpenAI(
base_url="https://api.grayrouter.com/v1",
api_key="YOUR_API_KEY",
)
response = client.chat.completions.create(
model="deepseek-v41-abliterated",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)02 / ACCESS CONTROL
Companies & API keys
| Company | Access | Active keys | Actions |
|---|
Your first company starts here.
Add Open English, Omnom, or another team, then issue their API keys.
Issued keys
Full keys are shown only once.| Key label | Company | Prefix | Created | Status |
|---|
03 / REPORTING
Usage by company & key
| Company / Key | Requests | Successes | Input tokens | Output tokens | Unmetered |
|---|
ESTIMATED GPU RUNTIME COSTSelected reporting period
—Total runtime
—Serving requests
—Startup
—Idle
Runtime estimates include unattributable usage. These are not provider invoices.
Compute sessions
Shared infrastructure · attributed to the starting company| Company | Provider | Started | Ready | Ended | Rate / hour |
|---|