What is Sansa?

Sansa is a simple AI API that gives you better performance at half the cost of using a single model provider. We intelligently route each request to the optimal AI model, so you get superior results without the premium price tag.

Why Developers Choose Sansa

Stop overpaying for simple requests. Not all requests are equal, some requests don't need the most expensive models. But how can you know ahead of time what a user will ask, and what model will best answer their query? You write evals, run tests. But not only is that costly, it is also blunt. You can use these results to help you determine what specific tasks (groups of tasks) should be handled by a specific model, but within those tasks, the variability is wide.

No more model evaluation hell. Tired of benchmarking models, building evaluation datasets, and maintaining routing logic? We've done the work. Our router is pre-trained on hundreds of thousands of real-world tasks - it already knows which models excel at what.

Actually works out of the box. Unlike competitors that demand weeks of setup and ML expertise, Sansa is three lines of code. Replace your existing SDK import, add your API key, and you're done. No configuration files, no learning period.

Performance you can trust. Our routing model has been trained on real production workloads, not just synthetic benchmarks. It's battle-tested, adaptive, and adds less latency than a database query.

How It Works

Router benefits. LLM API simplicity. That's Sansa.

Sansa uses a pre-trained routing model trained on hundreds of thousands of real-world examples. When you make a request to Sansa's API, our light weight router analyzes the request and routes to the optimal model in under 20ms.

It's as easy as using OpenAI or Anthropic. No complex setup, no new patterns to learn. Just point your existing code to Sansa.

Option 1: Use Your OpenAI SDK (Zero Changes)

Already using OpenAI? Just change the base URL, use your Sansa API key, and set the model to sansa-auto:

import OpenAI from 'openai';

const openai = new OpenAI({
  apiKey: process.env.SANSA_API_KEY,
  baseURL: 'https://api.sansa.ai/v1'
});

const result = await openai.chat.completions.create({
    model: 'sansa-auto',
    messages: [...],
});

Option 2: Use the Sansa SDK (More Visibility)

Get additional visibility into routing decisions and cost savings:

import { Sansa } from 'sansa';

const sansa = new Sansa({
  apiKey: process.env.SANSA_API_KEY,
});

const result = await sansa.create({
    callName: 'customer-support',
    model: 'sansa-auto',
    messages: [...],
});

That's it. Three lines of code. Better AI. Half the cost.

The Numbers

Performance: Sansa scores 96.78% on industry benchmarks, outperforming every single-model provider:

• Gemini 3 Pro: 89.8%
• Grok 4.1 Fast: 85.4%
• DeepSeek R1: 84.9%
• GPT-5.1: 80.1%
• Claude 4.5: 76%

Cost: 50% cheaper than frontier models on average.

What You Get

Better results than any single model. Because our router knows exactly which model is best at any given task, you get specialist-level performance across every request type. Our router is not a blunt instrument that routes "complex tasks" to large models and "simple tasks" to small ones. Our router is much more granular, we route to the exact model that will perform best. For example, in a customer service workload, we find that performance differs wildly based on the tone and sentiment of the customer. An angry customer looking for a refund is best handled by model A, while a calm customer looking for a refund is handled best by model B, and so on.

Immediate cost reduction. Most teams see 50-90% savings in their first month, depending on what model they are currently using, the more you are spending per token now, the more you will save. No ramp-up period, no learning curve - savings start with your first API call.

Your data stays private. We don't store or train on your data by default.

Built for Production

Proven performance. Our routing model scores 96.78% on industry benchmarks, outperforming every single-model provider.

Lightning fast. Sub-20ms routing decisions mean Sansa adds less latency than your typical database query. Your users won't notice the difference, but your CFO will notice the savings.

Secure by default. All models we route to are hosted in the US. We don't store or train on your data by default.

Who It's For

Anyone using AI APIs. Whether you're a startup making 100 requests per day or an enterprise processing millions, if you're paying for AI API calls, you're overpaying. Sansa gives you better AI at a lower cost. Bottom line.

Built for:

• Customer service chatbots
• AI agents
• API-driven AI applications
• Any product using LLM APIs