How Much AI Do We Actually Need?

I’ve been using almost exclusively open-weight models for AI-assisted software development, and while large models like GLM-5.3 are remarkably capable, their inference isn’t exactly cheap. So I’ve started experimenting with smaller, cheaper models to figure out where they’re good enough and where they fall short.

In this lightning talk, I want to share some observations and explore a few questions: How much model do we actually need for different tasks? When is a smaller model good enough, and when does it make sense to pay for a more capable one? And what does this mean for how we build software with AI? I don’t have definitive answers, but I’d like to share what I’ve learned so far and hear about your experiences as well.

Date

2026-11-12

Time

18:30 - 19:15

Conference / Event

INNOQ Technology Day Meetup Köln

Venue

innoQ Deutschland GmbH, Spichernstraße 44, 50672 Köln, Köln