# Mate Academy: $36K saved per year

Mate Academy, an online tech school, benchmarked models side by side on LLM API and moved each feature to the most cost-efficient one. It saves $36K per year on LLM spend. Implementation took one week.

Author: LLM API Team
Published: 2026-09-15
Updated: 2026-10-19
Canonical: https://llmapi.ai/case-study/mate-academy/

**Results**

- Saved per year: $36K
- From first call to production: 1 week

> “LLM API gave us the flexibility to benchmark models and choose the most cost-efficient option without compromising user experience.”
>
> — Mate Academy

## The challenge

Mate Academy runs AI features for its students. Each feature ran on one frontier model, so cost grew with every new cohort. Testing a cheaper model meant new integration work for each provider.

## What they did

The team moved every AI feature to LLM API and kept the OpenAI SDK. They benchmarked candidate models on real traffic. For each feature they picked the most cost-efficient model that kept the user experience the same. Automatic fallbacks send a request to a second provider when the first one fails.

## Why LLM API

- **One key for every model**: One OpenAI-compatible key, so switching models is a configuration change.
- **Per-feature cost analytics**: Models compared on real usage, feature by feature.
- **Lower prices**: Prices below provider list price on the models they kept.

Next step: [Talk to sales](https://meetings.hubspot.com/julia-andriienko/llmapi)