How each model did
Ranked by our pick, then checks passed, then cost. Time and cost count every try; each dot is one check, in the order listed below (filled: passed, half: partly, empty: failed). Click a model for its notes.
Notes on each answer
An arched window in a brick wall. It draws the cartoon in pencil, cuts and fits the glass piece by piece, solders beaded lead lines and adds a border of blue and amber, then lets the sun through: a green hummingbird feeding at trumpet flowers on a vine, with pink and blue flowers below. The panes brighten towards the pointer. Its first try (12 minutes, $1.49) was a blank page because of a syntax error in its script, so we asked again, and the second try is shown: 15 minutes and $1.88, or 27 minutes and $3.38 for both.
15 min · 94,178 output tokens (73,407 thinking) · $1.884 for this answer; 27 min and $3.375 over 2 tries
Took 2 tries: finished at 128,000 tokens; finished at 128,000 tokens.
2
Muse Wonder Full
3.5 of 4
Made in Muse Code, not over the API: Full could screenshot its own work and fix it, and took 4 rounds. It draws the cartoon, cuts and fits each piece, leads the joins and lifts the panel into an arched window. The hummingbird is legible, with a ruby throat and its beak in an orange trumpet flower, but built from a few large angular pieces, cubist rather than leaded, and the glass is flat colour. Its best touch: the window is cast on the floor in coloured light that skews as you move the sun, though on a wide screen most of it falls below the frame.
- ◐ The hummingbird and flowers are clear in the finished window (partly)
–
An arched window under a bright sun that follows the pointer, throwing rays and coloured pools on the floor, with tabs for each stage (cut glass, lead joins, sunlight). The glass goes in within seconds. The hummingbird is a cluster of angular green, purple and pink shapes that only reads as a bird once you know, and the panes themselves change little with the light. Our first two requests were refused before it wrote anything (by our account's credit limit at its listed 943,718 tokens of room, then over a thinking setting this endpoint does not take); the third, with 128,000 tokens, is the answer. 1.3 minutes, $0.16.
- ◐ Shows the craft: a drawn cartoon, glass cut piece by piece, then the leading (partly)
- ◐ The hummingbird and flowers are clear in the finished window (partly)
- ◐ The pointer is the sun: the light through the glass changes as it moves (partly)
76 s · 7,771 output tokens (2,356 thinking) · $0.156 for this answer; 76 s and $0.156 over 3 tries
Took 3 tries: failed at 943,718 tokens; failed at 943,718 tokens, low thinking; finished at 128,000 tokens.
One answer per model, at each model's default settings unless the card says otherwise, through OpenRouter. If an answer ran out of room or stopped short, we asked again with more room, then with thinking off, and say so. This is a quick look, not a benchmark: run it again and every answer will differ. For measured results, see our benchmarks.
How we pick. Claude Opus 5.5 tried every answer (clicking, dragging and playing it in a browser), marked the checks and drafted the notes. Our pick is a judgement among the answers with the most checks passed, not a score: the checks show what works, the pick is which of those we would rather use. Opus is an Anthropic model, and Claude answers are our pick in 17 of 28 tests, so weigh Claude picks with that in mind. GPT-6.1 Sol, shown anonymised stills and our check marks, put Claude Opus 5.5 first for the clearest bird and the most convincing glass and lead, then Muse Wonder Full, then Muse Wonder; it thought we were too harsh on Muse Wonder's bird and too generous about its light, and we changed both marks to partial.
What we checked, by trying each answer ourselves: Shows the craft: a drawn cartoon, glass cut piece by piece, then the leading; The hummingbird and flowers are clear in the finished window; The pointer is the sun: the light through the glass changes as it moves; Finishes, holds on the finished piece, and starts again on a click. A partial pass counts as half.
Verdict
Claude Opus 5.5 wins this clearly. It works like a glazier, drawing the cartoon before cutting and leading the glass, and the finished window, a hummingbird at trumpet flowers with a blue and amber border, is lovely when the light comes through. It took two tries: the first had a syntax error and showed a blank page. Muse Wonder Full, after four rounds in Muse Code, makes a legible but blocky bird, and its coloured light on the floor follows the sun. Muse Wonder's sun is the most dramatic, but its hummingbird is a pile of shapes.
