Low Tide: five models, one brief

The verdicts

Opus 5.5 came out on top of the first three. Sonnet 5.5 came close, for about 30% less money, and Sonnet 5 fell well behind both. Grok 4.7 and GPT 6.1 Sol are the same brief, shown beside those three. They don't have a place in that ranking.

Opus 5.5

1st · best style, best motion

The opening screen looks like a poster: a soft, oversized title over a sunset, with the type and the sun reflected in the water.
It has the most motion, and it all fits the dusk-on-the-beach mood.
A scrubber and play button let you step through the evening's schedule.
Of the Claude runs, it had the most careful accessibility work.

View Opus 5.5 ↑

Sonnet 5.5

2nd · very close, best value

It took its own direction, a light cream-and-navy look that feels like print. It doesn't copy the Opus design.
The typography is just as strong as Opus's, with a roman "Low" and an italic "Tide".
On its own it hit on the same idea as Opus: a slider you drag to move through the evening.
It has less motion than Opus, but the animations it does have are well chosen.

View Sonnet 5.5 ↑

Sonnet 5

3rd · good idea, weak execution

It has a real concept: film-reel framing with sprocket holes and sections numbered as reels.
The layout is a plain stack of boxes, and there's no image in the opening screen.
It has almost no motion.
It adds a live countdown to the gates opening, which is a useful touch.

View Sonnet 5 ↑

Grok 4.7

Evening slider, different type

The title is set in Anybody, a compressed sans, with a small sun beside the tagline. Body text is Literata. It also has a dark theme.
Drag the slider and the sun settles as the clock moves from the gates to the last film. The film that's on gets a highlight.
Buttons jump to the gate time, sunset, and each screening.
The FAQ is the three answers on the page. There is no accordion, and no keyframe animation.

View Grok 4.7 ↑

GPT 6.1 Sol

A printed poster

Huge Fraunces type, "Low" in roman and "Tide" in italic, a flat orange sun, and a thin wave along the bottom of the poster.
It picked Fraunces, same as the three Claude models. Body text is DM Sans.
The only motion is that wave, drifting slowly. It stops if you prefer reduced motion.
The FAQ opens and closes. There is no evening slider and no countdown.

View GPT 6.1 Sol ↑

The numbers

File size, animations and fonts are measured from the HTML. Cost, time and output tokens come from the Claude Code usage report, so those rows only cover the three Claude sessions.

	Opus 5.5	Sonnet 5.5	Sonnet 5	Grok 4.7	GPT 6.1 Sol
Cost	$1.94	$1.35	$1.08	—	—
Wall time	9 min	8 min	6 min	—	—
Output tokens	45.1k	57.8k	25k	—	—
File size	34 KB	23 KB	17 KB	19 KB	12 KB
CSS animations	9	4	1	0	1
Fonts	Fraunces, JetBrains Mono	Fraunces, DM Mono	Fraunces, IBM Plex Mono	Anybody, Literata, IBM Plex Mono	Fraunces, DM Sans, IBM Plex Mono

Sonnet 5.5 wrote the most code of the Claude sessions and still cost about 30% less than Opus 5.5.

Keep in mind

This was one prompt with one run per model. Another run could turn out differently.
The rankings are one person's taste. Look at the pages yourself and decide.
Opus 5.5, Sonnet 5.5, Sonnet 5 and GPT 6.1 Sol all picked Fraunces for the title. Grok 4.7 used Anybody and Literata.