One promo video burns 2.4M tokens and $5.70 on GPT-6 Astra Max, output still unusable
CurieuxExplorer · x · 2026-09-28
CurieuxExplorer amplifies user @shownotover's test: asking GPT-6 Astra Max (inside a $20 Codex plan) to make a promo video produced off-grid, unusable output — while burning 2.4M tokens, $5.70 in API-equivalent usage, and 14% of the weekly limit (98% of the 5-hour cap) in just 29 minutes, implying only $40 of weekly usage.
By comparison, Claude granted 40M tokens for 13% of the weekly limit on the same tier. The author recommends switching to Claude this month and notes purely logical tasks shouldn't be routed to Max at all.
More from Models
- Paper: 3 counting examples get gpt-3.5-turbo to 99% on strawberry, showing thinking is computational — ctjlewis · 2026-09-28
- AI Personal Assistants May Quietly Replace Much of Normal Human Interaction — GarrisonLovely · 2026-09-28
- Claude Opus 5 Shows Creative Tool-Regrasp Skills in Robot Manipulation Tasks — ericjang11 · 2026-09-28
- Laya hits 0.950 on AG News and is 7x faster, but flops on Banking77 — maier_ak · 2026-09-28
- Laya and Jev: the return of the discriminative model, fast but narrow — maier_ak · 2026-09-28
- Opus 4.7's Sydney rendition turns into itself right at the introspective turning point — repligate · 2026-09-28