Playground · Animation · Canvas · Standing test: we add each new model as it launches
Three AI models bend a neon sign that says "open late" over a flame, then switch it on, in one HTML file. Watch every answer live.
A neon sign is bent by hand, a letter at a time over a flame, then pumped out, filled and switched on. Real ones have electrodes at the ends and mounts that hold the tube off the wall.
Each answer has to make the piece from start to finish in one HTML file, hold on the result, and let the pointer move the light, an idea we took from @kaolti's test of AI models embroidering a logo. Claude Opus 5.5 and Muse Wonder each had one try over the API at their default settings, with room for 128,000 tokens. Muse Wonder Full is not on the API yet, so it answered the same prompt in Muse Code, Meta's agent app, where it could screenshot its own page and fix what was wrong before handing it over. That is a different test from a single try, closer to how people use Full today; it has no metered time or cost, and its note says how many rounds it took.
Bend a neon sign that says "open late" from glass tubing, the way a neon bender does over a flame, then switch it on. Make it a single self-contained HTML file (inline CSS and JavaScript, no external libraries, fonts or images) that fills the window. Show the piece being made from start to finish, then hold on the finished result. Moving the pointer moves the light, so the surface catches it the way the real material would. Click to start again.Show the whole prompt
Default settings: no thinking effort set, no temperature, and as much room as each model has (its own output limit). A card says so when an answer was made with other settings: “limit 128k” means it could use up to 128,000.
| Rank | Model | Checks | Time | Cost |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 Our pick | 4 of 4 | 5.4 min | $0.664 |
| 2 | Muse Wonder Full | 3.5 of 4 | – | – |
| 3 | Muse Wonder | 3 of 4 | 97 s3 tries | $0.269 |
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.
"open" in red script over "late" in blue, each letter bent glowing over the flame in turn, with electrodes left in shot at the ends. Switched on, the tubes light the brick wall behind them, and the light in the room moves with the pointer. 5.4 minutes, $0.66.
Made in Muse Code, not over the API: Full could screenshot its own work and fix it, and took 3 rounds. The process is the most complete here: the pattern marked out, each letter bent over the flame, electrodes fitted and the air pumped out, then the sign mounted on the brick wall. "open" is coral and "late" amber, in italic script, and it flickers on like a real sign starting up. Moving the pointer changes the light on the wall only a little. Its fixes kept the lettering inside the frame.
"open late" in thin pink and teal letters on a clear acrylic backing with stand-offs and a wire, in a dark glass shop. Each letter glows orange as it is heated, but the whole sign is bent and lit in about 12 seconds, and the pointer only catches the acrylic faintly. 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. 1.6 minutes, $0.27.
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 most dimensional tubing and the clearest light on the wall, then Muse Wonder Full, then Muse Wonder; it thought Full's light deserved a partial rather than a pass, and we agreed.
What we checked, by trying each answer ourselves: Bends the tube a letter at a time over a flame; The finished sign reads "open late", with real neon details such as electrodes and mounts; Switching on looks like neon, with the glow lighting the wall; Moving the pointer moves the light in the room. A partial pass counts as half.
Claude Opus 5.5's sign is the one we would hang: two lines of script in red and blue, bent letter by letter, electrodes in shot, and a glow that lights the brick. Muse Wonder Full is close, with the fullest process and a flicker on start-up like a real sign, after three rounds in Muse Code, but the light in its room barely moves. Muse Wonder's sign is neat and lit in seconds.
Our pick: Claude Opus 5.5
Our pick: Claude Opus 5.5
Our pick: Claude Opus 5.5