Why Google skips the frontier LLM race: cheap Flash models over beating rivals

burkov · x · 2026-09-17

Andriy Burkov argues Google isn't competing at the frontier because coding agents don't fit its business model. Casual users can't tell GPT-3.5-level AI Overviews from frontier models, so Google optimizes for making GPT-4.0-level quality cheap — hence Flash and Flash Light and their voice counterparts, cheap enough to be funded by ads. He claims Flash-class models are the best quality-per-dollar after extensive testing against open-weight alternatives. When a rival plateaus, Google can simply distill its best model. (Model names in the post are the author's speculation, unconfirmed.)

Original post →

More from Companies & People

Companies & People channel →