Start building
SandBase

Appearance

Model API Reference ​

Browse the models available through SandBase, find the exact model identifier to use, and copy a working API request.

Models by category ​

LLM Models ​

Use language and reasoning models for chat, tool calling, structured output, vision, and long-context tasks. Browse models by provider, including OpenAI, Anthropic, Google, DeepSeek, Alibaba, Meta, and ByteDance.

Image Generation ​

Generate, edit, transform, and upscale images through SandBase. Browse image models by provider, including OpenAI, Google, Ideogram, ByteDance, Recraft, Luma, and more.

Video Generation ​

Generate, edit, animate, upscale, and transform videos through SandBase. Browse video models by provider, including OpenAI, ByteDance, Kling, Google, Luma, Runway, Veed, and more.

Audio Generation ​

Generate, transform, clone, transcribe, and synthesize audio through SandBase. Browse audio models by provider and copy the exact async request format for each model.

Official Native API ​

Use provider-native protocols without rewriting an existing integration. Browse by provider and model; ByteDance Seedance uses the Volcengine Ark Contents Generations task protocol, while Google models use the Gemini-compatible GenerateContent or Interactions protocols.

Inference API contracts ​

The normalized request contracts are maintained in the linked API reference pages, while this section stays focused on choosing a model and its provider-specific schema:

Get started ​

Most LLM models use the OpenAI-compatible Chat Completions API. Anthropic models use the Anthropic Messages API; choose the protocol shown on the selected model page when both options are available.

bash
curl https://api.sandbase.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek/deepseek-v4-flash","messages":[{"role":"user","content":"Hello"}]}'

Image, video, and audio generation models use the unified /v1/run endpoint. The selected model's execution_mode determines whether the response is synchronous or returns 202 Accepted for polling:

bash
curl -X POST https://api.sandbase.ai/v1/run \
  -H "Authorization: Bearer sk-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-image-2","prompt":"A studio product photo on a clean white background"}'

When the response is 202 Accepted, poll the returned opaque run ID until the result is ready. Use the ID exactly as returned; do not add a task prefix or construct a provider-specific status path:

bash
curl https://api.sandbase.ai/v1/run/f3d2e8a1-7c4b-4a12-9d2e-123456789abc \
  -H "Authorization: Bearer sk-YOUR_KEY"