# LinkedIn post

OpenAI cut the price of GPT-6.1 Terra by about half on Oct 26. Prices like that now move every few weeks, at every lab.

Is your enterprise set up to take advantage? For most teams, switching models means a code change in every application that calls one, so a price cut sits unused for a quarter.

Metrum AI Router puts one endpoint in front of every lab and your own vLLM or SGLang pool. Applications call a model group by name. You add or swap the models behind that group with one config activation (no restart), and the applications don't change. Coming in the next release: the same swap as a single Admin API call. Run the routing policy in shadow on real traffic, evaluate cost and pass rate, then promote it or roll it back.

On our calculator, one developer's year of agent traffic costs $117,643 on a frontier-only setup and $17,646 routed.

It's Apache-2.0. Try the Terra switch on your own traffic this week.

## First comment

https://github.com/metrum-ai/router
https://github.com/metrum-ai/router/releases/tag/v4.3.0
https://github.com/metrum-ai/router/issues/395

## Public links

- Video: https://public-content.metrum.ai/metrum-ai-router/metrum-ai-router-token-prices-linkedin.mp4
- Captions: https://public-content.metrum.ai/metrum-ai-router/metrum-ai-router-token-prices-linkedin.vtt
- Thumbnail: https://public-content.metrum.ai/metrum-ai-router/metrum-ai-router-token-prices-linkedin-thumbnail.png
- Post text: https://public-content.metrum.ai/metrum-ai-router/metrum-ai-router-token-prices-linkedin-LINKEDIN.md

Upload the video natively to LinkedIn; these links are for sharing.
