Discussion on Model Performance in Long Contexts

dejavucoder · x · 2026-07-09

A discussion emerged regarding the performance of GPT-5.5 and Opus 4.8 at context lengths exceeding 100k–150k, with the original poster claiming the models become "dumber" with ultra-long contexts. A replier countered this, stating they can use models normally even when restricted to a 256k context.

Related event: Developers Debate LLM Long-Context Performance Limits(2 posts)→

Original post →

More from Models

Models channel →