Does 80% of a 256k context window degrade the same as 80% of 1M+?
BriefServe453 · reddit · 2026-09-11
A user asks whether retrieval and formatting accuracy at 80% utilization of a 256k window (200k tokens) matches 80% of a 1.05M window (840k tokens), i.e., whether context degradation scales by fill percentage or absolute token mass. Calls for benchmarks comparing deep retrieval at both scales.
More from Models
- User burns $200 of Codex credits in one agent turn — 4,700 of 5,000 credits, task unfinished — RileyRalmuto · 2026-09-11
- ChatGPT hit by outage in Europe: chat history vanishes, conversations won't load — doncorleonezzz · 2026-09-11
- Real token value of LLM subscriptions measured: SuperGrok Heavy hits 40x ROI, Cursor Ultra lowest — thesaraharminta · 2026-09-11
- After Navier-Stokes-level math results, even AI skeptics are now convinced AGI is here — cloneofsimo · 2026-09-11
- User ditches gpt-6-astra for gpt-5.6-sol, citing mood swings and absurd shortcuts — rudrank · 2026-09-11
- GPT Pro user puzzled why Codex 5.3 Spark gets separate usage tokens — takshit2 · 2026-09-11