Teknium: avoid 900k GPT variants — they charge extra beyond 272k context

Teknium · x · 2026-11-16

Responding to a Hermes Ultra subscriber complaining about fast usage burn, Nous Research's Teknium advises avoiding the 900k variants of GPT models, which incur extra charges beyond 272k context. He also says cache warning optimizations are coming soon that should save on all models.

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

Original post →

More from coding & agent

coding & agent channel →