The NotHoly API

Your models. Your own workflow.

Bring the published NotHoly models into your scripts, editor or agent stack. Use a familiar client, choose the model explicitly, and keep usage limits in your control.

Inference base URLhttps://api.notholy.com/v1
POST/v1/chat/completions
curl https://api.notholy.com/v1/chat/completions \
  -H "Authorization: Bearer $NOTHOLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "notholy-pulse",
  "messages": [
    {
      "role": "user",
      "content": "Review this endpoint for access-control flaws."
    }
  ],
  "max_tokens": 256
}'
Check field and protocol support

A documented contract. A deliberate request.

Choose the endpoint your client supports. Request fields, streaming events and model-specific differences are described in the API reference.

Compare published rates.

USD per million tokens. The effective request limits and supported fields are documented on each model page.

Scroll sideways to see all rates and limits.

ModelInputOutputCached inputContext
NotHoly SpecterGLM 5.3 Abliterated$3.00$6.00$0.451,048,576
NotHoly Pulse FlashGLM 5.3 Flash Abliterated$0.36$1.50$0.091,048,576
NotHoly Vector FlashDeepSeek V4.1 Flash 1017 Abliterated$0.42$1.50$0.1051,048,576

Make the limits explicit.

Choose the model.

Use the stable public ID. Each model page identifies the underlying model, effective request limits and current retail rates.

Set the budget.

Limit output and key spending before sending. Usage includes the context supplied and generated tokens, including billable reasoning where applicable.

Inspect the result.

Handle streaming, errors and tools according to the endpoint contract. An API call does not authorize arbitrary tool execution.

Review protocol and field support

Build from a bounded first call.

Create a customer key, review the published rates and follow the quickstart. Never place an API key in a public browser bundle.