Nous exec: avoid GPT 900k context variant — extra charges kick in past 272k

Teknium · x · 2026-11-16

Teknium of Nous Research warns users not to use the 900k context variant of GPT models, as extra charges apply beyond 272k context. He also teases upcoming "cache warming" optimizations that should save more on all models.

Related event: Nous' Teknium Warns Against GPT 900k Variant Over Extra Billing(3 posts)→

Original post →

More from Infra

Infra channel →