Models API

Every major AI model. One API key.

Use Claude, GPT, Gemini, Grok, DeepSeek and more through the OpenAI and Anthropic formats your code already speaks. Change the base URL and the key; keep your SDK.

Anthropic, OpenAI and Google models at the providers' own prices
Fallback to another model when a provider fails
Nothing stored by default

Start free Browse models

curl https://api.sparrowcodes.com/v1/chat/completions \
  -H "Authorization: Bearer $SPARROW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-5.5",
    "messages": [{"role": "user", "content": "Explain rate limiting in one paragraph."}]
  }'

Chat and coding models

Image models

Text to speech models

Transcription models

Embedding models

Reliability

Built for production traffic

The same models and the same API format. The difference is everything around the request.

Automatic fallback: List up to five backup models in models. If a provider errors or rate-limits, the next one answers, and you pay only for the answer.
Provider routing: Many models run on several providers. Sort by price, latency or throughput, pin an order, or block providers with provider.
Sparrow Auto: Send sparrow/auto and we pick a model for each request: fast ones for short prompts, stronger ones for code, tools and hard tasks.
Response cache: Identical requests answered from cache in milliseconds at no token cost, for up to 24 hours. Marked in the headers.
Spend limits per key: Lifetime or monthly limits on every key, plus alerts before your balance runs low.
Bring your own keys: Add your Anthropic, OpenAI or xAI key to use your own rates and quotas, with our routing and logs on top.

Fallback

When a provider fails, the next model answers

Errors caused by your request come straight back. Outages, rate limits and overloads move on to the next model, and the response headers tell you which one answered.

Fallback and routing docs

{
  "model": "anthropic/claude-sonnet-5.5",
  "models": ["openai/gpt-6.1-sol", "x-ai/grok-4.7"],
  "provider": { "sort": "price", "data_collection": "deny" },
  "messages": [{ "role": "user", "content": "…" }]
}

← X-Sparrow-Model: openai/gpt-6.1-sol
← X-Sparrow-Provider: openai

Endpoints

One base URL for chat, images, speech and more

OpenAI's Chat Completions and Responses formats and Anthropic's Messages format all work, on every chat model. Claude tools can run on GPT, and the other way round.

Chat models from Anthropic, DeepSeek, Google, OpenAI, qwen, xiaomi, z-ai.

Endpoint	What it does
/v1/chat/completions	Chat, tools, vision, streaming. OpenAI format.
/v1/messages	Claude's native format. Works with the Anthropic SDK and Claude Code.
/v1/responses	OpenAI Responses format, stateless.
/v1/embeddings	Text embeddings for search and RAG.
/v1/images/generations	Image generation.
/v1/audio/speech	Text to speech.
/v1/audio/transcriptions	Speech to text.
/v1/videos	Video generation, polled until ready.

Visibility

Know what every request cost

Cost on every response: usage.cost in US dollars on every response, and in the last streamed chunk.
Searchable activity log: Tokens, latency, provider, each fallback attempt and the exact cost, exportable to CSV.
Refused up front, not halfway: The worst-case cost is reserved before a request starts, so a request never stops mid-answer for lack of credit.
Webhooks and Slack: Low-balance, limit and key events in your own systems or a Slack channel.
Leaked key protection: If GitHub finds one of your keys in public code, we disable it within seconds and tell you.
Teams and roles: Owner, admin, developer and billing roles, two-factor sign-in and an audit log.

Keep going

Make your first request in minutes

Start free Read the docs