Jarbird blog.

Practical guides to measuring inference costs, routing requests and getting more from a prompt cache.

An egret flies low over a calm lake, with its reflection visible below.

When abliterated models are useful

Why models that refuse less often can help with defensive security, biotech documentation, and research. GLM-5.3 Flash Abliterated will run on Jarbird’s own infrastructure in the EU.

A bird flies past a coastal marker bearing the OpenAI logo.

How to reduce OpenAI API costs

Find out what a successful API call costs. Then try shorter prompts, prompt caching, and different routes to see what saves money.