Skip to main content
Sail provides the most efficient inference for autonomous AI agents. We serve trillions of tokens per week, with support for the best open-source models and your own LoRA fine-tunes.

​
Who should use Sail?

If:
  • You’re building an agent that largely acts on its own
  • You’re already running agents but face painful token costs, rate limits, or poor reliability
  • You’re looking for the most cost-efficient access to AI models
  • You have a token intensive, latency insensitive AI workload
…We’ve built Sail for you. Try us for free, starting with $1 in free credits.

Quickstart

Get an API key and make your first requests for free in minutes

​
The Sail difference

Our job is to dramatically increase intelligence per dollar, and make sure no compute goes to waste. We optimize the whole stack for throughput rather than single-turn latency. This tradeoff fits agentic work, delivering tokens reliably at a price that doesn’t limit the size of the tasks your agents can tackle. By leveraging Sail’s completion window options, you can get 2–5x more tokens per dollar compared to other open model APIs, and up to 10x more tokens per dollar compared to frontier model APIs.

​
API compatibility

Sail is OpenAI and Anthropic-compatible, with support for the OpenAI Chat Completions API, Open AI Responses API, and Anthropic Messages API. Migrating from a different provider is often as simple as specifying the Sail base URL, an API key, and a model name: 
import os

from openai import OpenAI

client = OpenAI(
    base_url="https://api.your-provider.com/v1",  
    base_url="https://api.sailresearch.com/v1",  
    api_key=os.environ["PROVIDER_API_KEY"],  
    api_key=os.environ["SAIL_API_KEY"],  
)

response = client.responses.create(
    model="<your-current-model>",  
    model="zai-org/GLM-5.1-FP8",  
    input="Explain the key ideas behind transformers.",
)
See Migrate to Sail for the full guide, including a ready-made prompt for running the migration with a coding agent, and the API support matrix for feature-by-feature coverage.

​
Maximize cost efficiency with Sail’s completion windows

Every request runs under a completion window, which specifies how long the workload can wait for a turn to finish. A longer window means a lower price for the same model and the same output. This gives you the flexibility to maximize cost efficiency in a way that’s appropriate for your workload.
WindowAvg. turn timePrice vs asap
asapImmediatebaseline
priority~1 min~30-50% lower
standard~5 min~45-65% lower
flexbest-effort~60-80% lower
Window availability varies by model. The Pricing page has the full support matrix.
Use the agent cost calculator to estimate the cost of running your agent with Sail inference.
  • Compared to other open-model providers: asap matches typical list prices, and the windows below it typically cost 2 to 5x less on supported models, with turnaround in minutes rather than a batch API’s 24-hour window.
  • Compared to a frontier API (GPT or Claude): a frontier-class open model at standard or flex typically costs 5 to 10x less for agentic workloads.

​
Rate limiting

Sail does not publish rate-limit tiers, and there is no limit-increase process to go through before scaling up. The service is designed to absorb large bursts of background traffic. In particular, choosing the flex completion window gives you access to the most relaxed rate limits available.

​
Next steps

Quickstart

Make your first API request!

Migrate to Sail

The three-line diff, or a prompt for your coding agent.

Models

Browse the catalog and pick your model.

Pricing

Per-token rates across every window.
Or email support@sailresearch.com to discuss exactly what you need.