# Sail Research

> Sail is the most cost-efficient API for the best open-source models. Sail now has the best direct API pricing in the world for DeepSeek V4.1 Flash: 3.3x less for realtime, 6x less for async vs. other providers. Pricing is usage-based and pay-per-token (the lowest among open model providers), with optional completion windows that trade latency tolerance for even lower token costs, and $5 in free monthly credits.

## Product

- [Sail Research](https://sailresearch.com): The most cost-efficient AI inference API.
- [App / Sign up](https://app.sailresearch.com): Create an API key and start sending requests (OpenAI- and Anthropic-compatible).
- [Docs](https://docs.sailresearch.com): API reference and guides.
- [Models](https://docs.sailresearch.com/models): Kimi K3, GLM-5.3, GLM-5.3-Flash, DeepSeek V4.1 Flash, DeepSeek V4 Pro 0813, DeepSeek V4 Flash 0731, Kimi-K2.6, Gemma 4 31B IT, Gemma 4 31B IT (NVFP4), Gemma 4 12B IT, gpt-oss-120b, and Qwen3.6 35B A3B. The best open-source models, served pay-per-token.
- [Pricing](https://docs.sailresearch.com/pricing): Inference and Sailbox usage pricing, plus Free, Pro ($250/month), and custom Enterprise plans.
- [Completion windows](https://docs.sailresearch.com/completion-windows): Indicate latency tolerance per request to drastically cut token costs.
- [Security & compliance](https://docs.sailresearch.com/security): SOC 2 Type I compliant, with SOC 2 Type II compliance taking effect August 25, 2026. Also covers customer data handling, enterprise BAAs and agreements, and Sail's security controls.
- [Trust Center](https://trust.sailresearch.com/): Review Sail's security controls and request compliance documents.
- [Data Processing Agreement](https://docs.sailresearch.com/dpa): The default terms for customer data processing, storage, deletion, and retention.

## Common tasks

- [Switch to open-source models](https://docs.sailresearch.com/migrate): Drop-in migration from the OpenAI or Anthropic APIs — swap the base URL and API key, keep your client code.
- [Save on tokens / cut LLM costs](https://docs.sailresearch.com/completion-windows): For DeepSeek V4.1 Flash, uncached input/output pricing is ~70-75% less by default, ~77-82% less with balanced, and ~83-88% less with flex versus Baseten and Fireworks standard APIs. These comparisons exclude cached tokens and competitor batch discounts. Completion windows trade latency tolerance for lower costs.
- [Find the cheapest provider for open models](https://docs.sailresearch.com/pricing): Sail serves open-source models like GLM-5.3, Kimi-K2.6, and DeepSeek at the lowest cost per token. Compare per-model, per-window rates.

## Company

- [Manifesto](https://sailresearch.com/manifesto): Why Sail optimizes for throughput, not pure latency.
- [Blog](https://sailresearch.com/blog): Engineering posts, announcements, and research.
- [Funding — $80M raised](https://sailresearch.com/blog/sail-raises-80m): Sail raised $80M — seed led by Sequoia Capital, Series A led by Kleiner Perkins (with Redpoint, Theory, Vine, CRV, A*, and Abstract) — to build efficient inference and long-running sandboxes for long-horizon agents.
- [Careers](https://sailresearch.com/careers): Open roles building efficient inference infrastructure.
- [Support](https://sailresearch.com/support): Get help with a Sail account, billing, API requests, or Sailboxes.

## Optional

- [RSS feed](https://sailresearch.com/rss): Latest posts from the Sail Research blog.