GeniusRelay
EN中文Sign up

The unified API
for AI models

Access 599 models from 37 providers through a single OpenAI-compatible endpoint, including uncensored editions of GLM, DeepSeek, Qwen and MiMo. Prepaid, usage-based pricing.

Get API keyapi.geniusrelaygpt.com/v1
599
models
37
providers
43
uncensored
2M
max context
  • Cursor
  • Cline
  • SillyTavern
  • Codex CLI
  • Cherry Studio
  • LobeChat
  • Chatbox
  • OpenAI SDK
  • Continue
  • Roo Code
  • Open WebUI
  • NextChat

One API for every model

599 text models from 37 providers. 333 accept images and 404 support reasoning.

Browse all models
DeepSeek V4.1 Flash Hereticdeepseek-v4.1-flash-heretic$0.271/M input512K contextApex Flash 1 Abliteratedglm-5.3-flash-abliterated$0.205/M input1M contextDeepSeek V4.1 Flash Uncensoreddeepseek-v4.1-flash-uncensored$0.452/M input1M contextMiMo V2.6 Flash Uncensoredmimo-v2.6-flash-uncensored$0.685/M input1M contextGLM 5.3 Flash Uncensoredglm-5.3-flash-uncensored$0.274/M input1M contextMuse Embermuse-ember$0.137/M input1.05M contextGemini 4 Carbongemini-4-carbon$2.74/M input1.05M contextERNIE Light DT 1.0 (Intl)ernie-light-dt-1.0-intl$1.78/M input1.02M contextNex N2.6nex-n2.6$0.205/M input262K context

View all 599 models

Switch models by changing one parameter

Every model shares one base URL, one key and one request format. To move from GLM to DeepSeek, change the model field in the request body. Nothing else in your code changes.

API reference
POST api.geniusrelaygpt.com/v1/chat/completions
{
  "model": "deepseek-v4.1-flash-heretic",
  "messages": [{ "role": "user", "content": "…" }],
  "stream": true
}
Context
512K
Input /M
$0.271
Output /M
$1.08

Automatic failover across routes

A model can be served over several routes. When one returns an error or is rate limited, the gateway retries the next route in priority order. You do not need retry logic in your code, and failed requests are not billed.

Errors and retries
Route statusSample

Route A

Route B

Route C

Through GeniusRelay

Usage and spend in one console

Every request is logged with its model, input and output tokens and cost. Filter by time range and export the log for reconciliation.

Open the console
UsageSample · last 14 days
Requests
12,480
Token
84.2M
Spend
$37.60
  • DeepSeek V4.1 Flash Heretic
  • Apex Flash 1 Abliterated
  • DeepSeek V4.1 Flash Uncensored
  • Other

Key spending limit$37.60 / $50.00

When the limit is reached this key stops working. Other keys are unaffected.

Per-key controls

Issue separate keys for each project or teammate and limit how much each can spend, when it expires and which models it can call. If a key leaks, revoke it without affecting the others.

  • Unlimited keys
  • Spending limits
  • Expiry dates
  • Model allowlists
  • IP allowlists
  • Usage log export
API key · sk-prod-****8f2aSample
Spending limit$50.00
Expires2027-01-31
IP allowlist203.0.113.0/24
Allowed modelsdeepseek-v4.1-flash-hereticglm-5.3-flash-abliterateddeepseek-v4.1-flash-uncensored

What people build

Choose models by use case. Each comes with a setup guide.

2M context

Long-form writing and document processing

384 models offer a context window of 256K or more, up to 2M, enough for a full manuscript or a very long conversation.

View modelsPython example

Send your first request in three steps

  1. Sign upCreate an account in the console.
  2. Add creditsRedeem a top-up code under Wallet.
  3. Call the APICreate a key under API Keys and use it in the code below.
curl https://api.geniusrelaygpt.com/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-heretic",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Estimate your cost

Figures use current list prices. The lowest-priced model is Ling 3.0 Tiny at $0.0206 per million input tokens. Raising the cached-input share lowers the cost of long conversations.

Model
Per request
$0.00168
Per day
$0.8410
Per month (30 days)
$25.23

DeepSeek V4.1 Flash Heretic: input $0.2713/M · output $1.085/M · cached input $0.0181/M

Request with these settings
curl https://api.geniusrelaygpt.com/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-v4.1-flash-heretic", "max_tokens": 800,
       "messages": [{"role": "user", "content": "…"}]}'

37 providers

Select a provider to see its models.

Technical overview

Base URL
https://api.geniusrelaygpt.com/v1
Authentication
Header Authorization: Bearer sk-…. Keys are created under API Keys in the console.
Endpoints
POST /chat/completions (streaming and non-streaming), GET /models
Billing
Input, output and cached input tokens are priced separately. The cost of each request appears in Usage Logs.
Payment
Prepaid credits, added by redeeming a top-up code in the console.
Out of credits
Requests return 402 with insufficient_quota. You are never billed in arrears, and each key can have its own spending limit.
Error format
{"error": {"message", "code", "request_id"}}. Quote the request_id when contacting support.
Integrations
Cursor, Cline, SillyTavern, Codex CLI and the Python SDK each have a step-by-step guide.

Frequently asked questions

How is this different from calling a model provider directly?

The request format is the same (OpenAI-compatible). The difference is that one key reaches every model, billing is consolidated, and uncensored editions are available.

How do I add credits?

Purchase a top-up code and redeem it under Wallet in the console. Available denominations are listed at checkout.

Can I be charged more than I expect?

No. Billing is prepaid, so requests are rejected once credits run out. You can also set a spending limit and an expiry date on each key.

Are streaming and tool calling supported?

Streaming (SSE) is supported for every model. Tool calling depends on the model: 429 models support it, and the Models page has a filter for it.

Are there restrictions on uncensored models?

Yes. Sexual content involving minors is prohibited, as is use for fraud, cyberattacks or other illegal activity. See the Terms of Service. Request logs are retained.

How do I troubleshoot a failed request?

The response body contains a message, a code and a request_id. The Errors page explains each code. Include the request_id when you contact support.

More questions in the docs

One API key, 599 models.

Create an account, add credits and start calling. Pay as you go, no subscription.

Sign up freeQuickstart