Choosing a Model

How to pick and switch between AI models on Archipelag.io — LLMs, image generators, speech, and more

Choosing a Model

Archipelag.io offers 130+ Cargos across LLMs, image generators, speech-to-text, embeddings, and more. Every interface defaults to the best available model — but you can switch to a specific one whenever you want.

Quick Start (Zero Config)

Just open the chat or image generation page and start typing. The platform automatically picks the best model based on what Islands are currently online:

• Chat → defaults to the smallest fast LLM (Qwen3.5-0.8B or Mistral 7B)
• Image generation → defaults to FLUX.1-schnell or Stable Diffusion
• API → defaults to the best LLM when you use model: "gpt-3.5-turbo" or any OpenAI model name

Model Picker (Chat)

Above the chat input, there’s a compact Model dropdown:

• Auto — lets the platform pick the best available model (default)
• Specific models — choose from all available LLMs

Available LLM models include:

Model Picker (Image Generation)

The image generation page has the same dropdown pattern:

Video generation models (Wan2.1, HunyuanVideo, CogVideoX) are available through the API.

API — Choosing a Model

OpenAI-Compatible Endpoint

Use standard OpenAI model names — they map to the best available Archipelag.io model:

curl -X POST https://app.archipelag.io/api/v1/chat/completions \
  -H "Authorization: Bearer ak_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-3.5-turbo",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Model Name Mapping

Using Cargo Slugs Directly

You can also use any Cargo slug as the model name:

# Use a specific GGUF model
curl -X POST https://app.archipelag.io/api/v1/chat/completions \
  -H "Authorization: Bearer ak_YOUR_KEY" \
  -d '{"model": "gguf-llama-3.1-8b", "messages": [...]}'

# Use the native Mistral model
curl -X POST https://app.archipelag.io/api/v1/chat/completions \
  -d '{"model": "gguf-mistral-7b", "messages": [...]}'

Short aliases also work: mistral, qwen-4b, llama-8b.

Unknown Model Names

If you pass a model name that doesn’t match any mapping or Cargo slug, the API automatically selects the best available LLM instead of returning an error. This makes it safe to use Archipelag.io as a drop-in replacement for OpenAI.

Catalog API

Browse all available Cargos programmatically:

# All Cargos
curl https://app.archipelag.io/api/v1/cargos

# Just LLMs
curl https://app.archipelag.io/api/v1/cargos?category=llm

# Just image generators
curl https://app.archipelag.io/api/v1/cargos?category=image

# Filter by runtime type
curl https://app.archipelag.io/api/v1/cargos?runtime_type=onnx

Categories

Response Format

{
  "cargos": [
    {
      "slug": "gguf-mistral-7b",
      "name": "Mistral 7B Instruct",
      "description": "Mistral 7B Instruct v0.2 (Q4_K_M)",
      "runtime_type": "llmcpp",
      "onnx_task_type": null,
      "price_per_job": 1.0,
      "required_vram_mb": 6000,
      "required_ram_mb": 8000
    }
  ],
  "count": 15
}

Jobs API — Any Cargo

For non-chat Cargos (ONNX, WASM, containers), use the Jobs API directly:

# Submit a sentiment analysis job
curl -X POST https://app.archipelag.io/api/v1/jobs \
  -H "Authorization: Bearer ak_YOUR_KEY" \
  -d '{
    "workload": "onnx-sentiment",
    "input": {"text": "This product is amazing!"}
  }'

# Submit an object detection job
curl -X POST https://app.archipelag.io/api/v1/jobs \
  -d '{
    "workload": "onnx-detect",
    "input": {"image": "<base64 encoded image>", "threshold": 0.5}
  }'

Next Steps

{% end %}