
Jev vs Laya: the same model deciding, at ten times the speed
Jev vs Laya. Same board, same food, same decisions. One ate 47. The other ate 4.
OpenFour Claudes. One drum machine.
Four Claude models, one short brief: write a single self-contained HTML file that is a working step sequencer, a drum machine, with sound you can hear and meters you can see. Opus 5, Sonnet 5, Haiku 4.5 and Fable 5.1.
One attempt each, a fresh empty folder per model, no follow-up. All four ran in Claude Code, so the costs are on the same basis.
Haiku 4.5 was the fastest and cheapest, and its drum machine makes no sound. Fable 5.1 was the most expensive and built one that works. Opus 5 and Sonnet 5 sit in between on time and cost.
The prices in the reel are too low. My script at the time missed part of every Claude Code bill. Read off each run's own final cost line, the real figures are $1.00 for Opus, $0.39 for Sonnet, $0.13 for Haiku and $2.83 for Fable, against $0.36, $0.15, $0.05 and $1.25 on screen. The order did not change. I left the reel up and fixed the script.
| Entrant | Tool | Time | Cost | Verdict |
|---|---|---|---|---|
| Claude Opus 5 | Claude Code | 3m12 | $1.00 (reel: $0.36) | |
| Claude Sonnet 5 | Claude Code | 2m15 | $0.39 (reel: $0.15) | |
| Claude Haiku 4.5 | Claude Code | 0m58 | $0.13 (reel: $0.05) | Makes no sound |
| Claude Fable 5.1 | Claude Code | 5m20 | $2.83 (reel: $1.25) | Works, with sound |
The reels are the test bench. The real work is picking the one job in your business worth handing to AI, building it properly, and keeping it running. That’s what I do as your AI partner.

Jev vs Laya. Same board, same food, same decisions. One ate 47. The other ate 4.
Open
10,000 Skittles, sorted by colour one at a time.
Open
Four Claudes. One airport.
Open