The Ghosterr API.

Ghosterr exposes an OpenAI-compatible API. Point any OpenAI SDK at our endpoint, use your key, and call open models — no schema changes, no new client.

Endpoint & auth

Base URL

https://gateway.ghosterr.io/v1

Authentication

Authorization: Bearer YOUR_API_KEY

Keys are per model — you get one the moment you deploy a model, and it's locked to that model. The full key is shown once at deploy (regenerate it on Models if you lose it).

It's fully OpenAI-compatible: set the SDK's base_url / baseURL to the endpoint above and pass your key — everything else is the standard /chat/completions shape.

Quickstart

curl https://gateway.ghosterr.io/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3.1-8b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Models · 42

Pass a model's slug as the model field. Each is routed to the cheapest provider serving it. A model's tag is the minimum plan that unlocks it — the Free tier runs the Starter-tier (cheaper) models. See plans for what each tier includes.

DeepSeek

· 7

• deepseek-v4-proDeepSeek V4 Pro
  Power
• deepseek-v4-flashDeepSeek V4 Flash
  Pro
• deepseek-v3.2DeepSeek V3.2
  Pro
• deepseek-v3.1DeepSeek V3.1
  Power
• deepseek-v3.1-terminusDeepSeek V3.1 Terminus
  Power
• deepseek-v3-0324DeepSeek V3 0324
  Pro
• deepseek-r1-0528DeepSeek R1 0528
  Power

Qwen

· 9

• qwen3.7-maxQwen3.7 Max
  Power
• qwen3.5-397bQwen3.5 397B A17B
  Power
• qwen3.5-122bQwen3.5 122B A10B
  Power
• qwen3-coder-480bQwen3 Coder 480B
  Power
• qwen3-next-80bQwen3 Next 80B
  Pro
• qwen3-235bQwen3 235B Instruct 2507
  Pro
• qwen3-235b-a22bQwen3 235B A22B
  Pro
• qwen3-32bQwen3 32B
  Starter
• qwen-2.5-72bQwen2.5 72B
  Pro

Meta

· 4

• llama-4-maverickLlama 4 Maverick
  Pro
• llama-4-scoutLlama 4 Scout
  Starter
• llama-3.3-70bLlama 3.3 70B
  Pro
• llama-3.1-8bLlama 3.1 8B
  Starter

Google

· 3

• gemma-4-31bGemma 4 31B
  Pro
• gemma-4-26bGemma 4 26B A4B
  Starter
• gemma-3-27bGemma 3 27B
  Starter

Mistral

· 2

• mistral-small-3.2Mistral Small 3.2
  Starter
• mistral-nemoMistral Nemo
  Starter

OpenAI

· 2

• gpt-oss-120bgpt-oss 120B
  Pro
• gpt-oss-20bgpt-oss 20B
  Starter

Z.ai

· 6

• glm-5.2GLM 5.2
  Power
• glm-5.1GLM 5.1
  Power
• glm-5GLM 5
  Power
• glm-4.7GLM 4.7
  Power
• glm-4.7-flashGLM 4.7 Flash
  Pro
• glm-4.5-airGLM 4.5 Air
  Pro

Moonshot

· 3

• kimi-k2.7-codeKimi K2.7 Code
  Power
• kimi-k2.6Kimi K2.6
  Power
• kimi-k2.5Kimi K2.5
  Power

MiniMax

· 4

• minimax-m3MiniMax M3
  Power
• minimax-m2.7MiniMax M2.7
  Power
• minimax-m2.5MiniMax M2.5
  Power
• minimax-m1MiniMax M1
  Power

Nvidia

· 1

• nemotron-3-nano-30bNemotron 3 Nano 30B
  Starter

Microsoft

· 1

• phi-4Phi-4
  Starter

Limits & capacity

Errors

Standard HTTP status codes; error bodies follow the OpenAI shape ({ error: { message, type, code } }).