Video cover

Theo assesses OpenAI's October 29 usage-limit increase through three lenses: API prices, cost per completed task, and subscription allowances. This opinion/news video combines reported pricing charts with his own blind code-review comparison to explain why he still favors Opus 5.5 for daily coding. All prices, plan rules, and judgments describe the video at recording time. 0:24 7:34 12:00 14:40

Topic breakdown

  1. Opening argument and Passport sponsor segment
  2. What the usage-limit increase changes
  3. Token prices versus cost per task
  4. Subscription-adjusted value and the capability frontier
  5. Why usage resets change the comparison
  6. Blind code-review results and numerical caveats
  7. Shared limits and Theo's final recommendation

Key takeaways

  • The $200 OpenAI plan's doubling restores an earlier allowance cut; the $100 and $500 plans receive increases without that earlier cut, according to Theo. 4:11 15:01
  • API-equivalent allowance alone favors Claude in his figures: about $12,000 on its $200 plan versus $4,300 on OpenAI's $200 plan. Cheaper OpenAI tasks narrow that difference in actual work delivered. 5:03 5:33
  • Subscription-adjusted benchmark tasks cost approximately $0.05 for Opus 5.5 high and $0.04 for 6.1 Soul high in the cited analysis. These are benchmark estimates, not the costs of Theo's code reviews. 8:01
  • Both model judges preferred Opus reviews in most of 83 comparisons: Opus chose it 66 times and Soul chose it 54 times. 12:00
  • Theo reports roughly 205 Opus reviews versus 290 Soul reviews per subscription, favoring Opus for quality and Soul for throughput. The quoted cost and allowance figures do not reproduce those counts; see the numerical caveats. 12:26 12:58