# FasterTokens

> FasterTokens is an OpenAI-compatible endpoint for voice agents. For every turn of DeepSeek V4.1 Flash it queries multiple inference providers simultaneously and optimizes for time to first token, so one slow provider never becomes dead air on a call. Latency is measured around the clock and published.

OpenAI-compatible: POST https://api.fastertokens.com/v1/chat/completions with model "deepseek-v4.1-flash" and an API key as a Bearer token. Streaming and tool calls are supported. For every request it queries multiple inference providers simultaneously and optimizes for time to first token. Live benchmark data (JSON, no key): https://api.fastertokens.com/benchmark

## Docs

- [Quickstart](https://fastertokens.com/docs.md): Get a key and send your first completion in a few minutes.
- [Voice agent setup](https://fastertokens.com/docs/voice-agents.md): Point a voice agent framework or platform at FasterTokens, with a fallback model and one session per call.
- [API reference](https://fastertokens.com/docs/api-reference.md): Endpoints, headers, errors and limits for the FasterTokens API.

## Optional

- [Full docs in one file](https://fastertokens.com/llms-full.txt)
- [Benchmark](https://fastertokens.com/benchmark)
- [Privacy](https://fastertokens.com/privacy)