Sonnet 5.5 did this. Opus 5.5 quality with half price.
A 30-second, 1080p, 60 fps video featuring Snoo, kinetic type, a subreddit marquee, and various Reddit-themed animations was created using Sonnet 5.5. The project involved 14 sub-agents and consumed 739k tokens, 101.6M cache reads, and 3.07M cache writes, costing $35.40 with Sonnet 5.5. This was approximately 43% cheaper than if Opus 5.5 had been used for the same tokens, primarily because cache reads, which constitute most of the token usage, cost the same for both models.
This report uniquely details the cost breakdown, showing Sonnet 5.5 achieved Opus 5.5-level quality for 43% less due to cache read pricing, unlike general performance claims.
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年9月29日 13:42 UTC
收录当时偏移：UTC+02026年9月29日 16:00 UTC
- 发布
- 2026年9月29日 13:42
- 收录
- 2026年9月29日 16:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 同步延迟
档位是按信源手工设定的编辑判断，不是逐条打分。
讨论趋势
百分比基于采集到的讨论信号，不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。
It's 30 s, 1080p, 60 fps. Snoo assembles itself, kinetic type, a subreddit marquee, a 3D upvote with a karma counter, an isometric feed that sorts itself, an r\place timelapse, a trending bar-chart race, feature tiles, a comment thread that turns into Snoo, a TIL / ELI5 / AMA montage, and an end card.
Everything is code: canvas frames drawn in headless Chrome and encoded with ffmpeg, plus a synthesized soundtrack. No After Effects, no stock footage, no video model.
What it cost. I'm on the Max 20x plan, so I never saw a bill. These are API list-price equivalents that I worked out from the session transcripts.
Output tokens Cache reads Cache writes At Sonnet 5.5 prices Same tokens at Opus 5.5 prices 15 s version (one chat) 210k 10.6M 0.29M $4.94 $7.77 30 s version (14 sub-agents) 529k 91.0M 2.79M $30.46 $42.72 Total 739k 101.6M 3.07M $35.40 $50.48 Prices per million tokens: Sonnet 5.5 is $2 in / $10 out, Opus 5.5 is $4 / $20, and cache reads are $0.20 on both.
Cache writes I assumed at 1.25x input. The whole thing was 779 tool calls (161 in the main chat, 618 by the sub-agents) over about 2.4 hours, and much of that was waiting on renders.
So on paper Opus 5.5 would have been about 43% more expensive for the same tokens. It's not 2x because almost all the tokens are cache reads, which cost the same on both models.