The answers to “Pour latte art into a cup of coffee”, side by side

Playground · Animation · Canvas · Standing test: we add each new model as it launches

Pour latte art into a cup of coffee

AI models pour latte art: a tulip from a milk jug into espresso, seen from above, in one HTML file. Few make a tulip. Every answer runs live.

About this test

A tulip is a standard latte-art pattern: a few pushed hearts stacked up and cut through with a final pull. It is easy to check, and easy to get wrong as a heart or a leaf.

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.

Read more

The prompt, word for word

Pour latte art into a cup of coffee: a tulip, poured from a milk jug into the espresso, seen from above.

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.

Each answer makes the piece once, then holds on it. Move the pointer to move the light; click to start again.

How each model did

RankModelChecksTimeCost
1 GPT-6.1 Terra Our pick 3.5 of 4 72 s $0.032
2 Muse Wonder Full 3.5 of 4 – –
3 Kimi K3.1 Pro 3.5 of 4 19 min $0.408
4 MiniMax M3.1 Flash (free route) 3 of 4 8.7 min $0.025
5 Claude Opus 5.5 3 of 4 6.6 min $0.784
6 DeepSeek V4.1 Flash 1017 2.5 of 4 2.3 min $0.045
7 Muse Wonder 2 of 4 67 s3 tries $0.178
8 Mercury 3 0.5 of 4 5.2 min2 tries $0.002
9 Gemini 4 Argon Pending 89 s $0.135

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.

What each answer cost, against the checks it passed

Best value: GPT-6.1 Terra, 3.5 of 4 checks for $0.032. The best value is the cheapest answer with the most checks passed, counting every try.

0%20%40%60%80%100% $0.001$0.01$0.1$1 Cost, every try (log scale) Checks passed GPT-6.1 Terra

Best valueOther answers

Notes on each answer

1

GPT-6.1 Terra

Our pick 3.5 of 4

A cup from above and a steel jug that pours a thin stream: a base heart, then hearts pushed up one after another, and a pull-through from top to tail. The finished tulip has four clear tiers and a fine stem and fills the cup well, the cleanest pattern on this page. The light moves over the surface and the cup. The cup is neat but illustrated, and the crema is a smooth, even brown rather than foam. It holds and pours again on a click. 72 seconds, $0.032.

  • The crema and the cup look real (partly)

72 s · 6,454 output tokens (1,017 thinking) · $0.032

Added 27 October 2026.
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 3 rounds. The only tulip of the first six answers: the espresso settles, the milk builds a base, five petals are pushed in one after another, and a pull-through cuts the stack, leaving a small heart on top. It is small in the cup and still close to a leaf if you are strict, and the crema is dotted with large spots, so the cup looks less real than Opus's. The light glints across the surface. Its fixes made the hearts rounder and wider apart so that it read as a tulip.

  • The crema and the cup look real (partly)

–

3

Kimi K3.1 Pro

3.5 of 4

A jug swings in over a cup on a saucer and pours a growing white mass, then a small heart and a thin pull-through. The cup and the swirled crema look real, and the light moves across them. The pattern does not convince: wide tiers with a heart on top and a short stem, more a fir tree than a tulip. It holds and pours again on a click. 19 minutes, $0.41.

  • The finished pattern is a tulip: stacked petals with a pull-through (partly)

19 min · 40,731 output tokens (29,138 thinking) · $0.408

Added 27 October 2026.
4

MiniMax M3.1 Flash (free route)

3 of 4

A white cup from above with a dark crema and a silver jug that tips in and pours a thin stream, finished in about ten seconds; it holds and pours again on a click. The pattern is a small outline: two nested curved petals with a line pulled through and a short stem, more like a pale onion than a tulip's stacked, filled petals. The crema is a smooth gradient and the cup is plain, and the highlight on the rim and the coffee moves with the pointer. One try, 8.7 minutes on OpenRouter's free route, so the time is not representative; $0.025 at list prices.

  • The finished pattern is a tulip: stacked petals with a pull-through (partly)
  • The crema and the cup look real (partly)

8.7 min · 41,759 output tokens (29,628 thinking) · $0.025

Added 19 October 2026.
5

Claude Opus 5.5

3 of 4

The most realistic cup: a white cup and saucer on dark wood, a crema with real texture, and a steel jug pouring from below. But the milk lands as a lumpy white column with jagged edges, not a tulip, and it sits flat on the crema. The light moves across the cup rim and the crema. 6.6 minutes, $0.78.

  • The finished pattern is a tulip: stacked petals with a pull-through

6.6 min · 39,176 output tokens (26,407 thinking) · $0.784

6

DeepSeek V4.1 Flash 1017

2.5 of 4

A white cup seen from above with a smooth brown crema. There is no jug: the milk is a bright dot landing near the rim, and the pattern grows in about four seconds. It ends on four stacked cream arcs with a pull-through and a stem, the right structure for a tulip, but flat and opaque, so it looks more like a cartoon fir tree than milk foam. The highlight moves across the crema and the rim with the pointer. 2.3 minutes, $0.045.

  • Shows the pour from the jug, with the milk landing in the espresso (partly)
  • The finished pattern is a tulip: stacked petals with a pull-through (partly)
  • The crema and the cup look real (partly)

2.3 min · 37,676 output tokens (31,100 thinking) · $0.045

Added 19 October 2026.
7

Muse Wonder

2 of 4

Titled "Tulip", with steam rising and the jug pouring from above, but it pours a large leaf, a rosetta with a centre line and nested chevrons. Clean and pretty, and the wrong pattern. 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. 1.1 minutes, $0.18.

  • The finished pattern is a tulip: stacked petals with a pull-through
  • The crema and the cup look real (partly)
  • The light follows the pointer across the surface (partly)

67 s · 8,868 output tokens (3,274 thinking) · $0.178 for this answer; 67 s and $0.178 over 3 tries

Took 3 tries: failed at 943,718 tokens; failed at 943,718 tokens, low thinking; finished at 128,000 tokens.
8

Mercury 3

0.5 of 4

A dark cup seen at a low angle, with a small cream cloud on a stalk, like a tree or a mushroom, in the middle of the coffee, finished within a few seconds. There is no jug and no pour, the pattern is not a tulip, the cup is a thin ring with a flat brown surface and a faint grey box behind the pattern, and moving the pointer barely changes the light. Two tries: the first stopped unfinished after five minutes, and the second, with thinking off, took 9 seconds. 5.2 minutes and $0.0023 in all.

Missed or only partly met 4 of the 4 checks.

9 s · 4,712 output tokens · $0.002 for this answer; 5.2 min and $0.002 over 2 tries

Took 2 tries: stopped mid-page at 131,072 tokens; finished at 131,072 tokens, thinking off.Added 19 October 2026.

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 29 tests, so weigh Claude picks with that in mind. Before GPT-6.1 Terra was added, GPT-6.1 Sol, shown anonymised stills and our check marks, put Muse Wonder Full first as the only recognisable tulip, then Claude Opus 5.5 for its cup and crema, then Muse Wonder, whose pattern it read as a leaf. We agreed at the time. Terra's pick has no second opinion yet.

What we checked, by trying each answer ourselves: Shows the pour from the jug, with the milk landing in the espresso; The finished pattern is a tulip: stacked petals with a pull-through; The crema and the cup look real; The light follows the pointer across the surface. A partial pass counts as half.

Verdict

GPT-6.1 Terra, added on 27 October, is our pick: it pours the clearest tulip here, four stacked hearts with a clean pull-through, from a jug, in one try, in 72 seconds for $0.032. Its crema is flat next to Claude Opus 5.5's, but the pattern is the test. Muse Wonder Full was our pick until then and made the only tulip of the first six answers, stacked petals with a pull-through, in three rounds in Muse Code, where it looked at its own page and reshaped the hearts; its tulip is smaller and its crema is spotted. Claude Opus 5.5 has by far the most realistic cup and crema, but its milk lands as a lumpy blob, and Muse Wonder poured a leaf. DeepSeek V4.1 Flash 1017, added on 19 October, comes closer than Opus: stacked arcs with a pull-through, but flat, and poured with no jug. MiniMax M3.1 Flash, added the same day on OpenRouter's free route, pours from a jug, but its pattern is a small outline rather than a tulip. Mercury 3, added the same day, needed two tries and ends on a small cream cloud on a stalk, with no jug. Kimi K3.1 Pro, added on 27 October, has a realistic cup and a real pour, but its wide tiers look more like a fir tree than a tulip.

What would your prompt get?The Spring Prompt playground runs one prompt on several models at once, from OpenAI, Anthropic, Google, Mistral and others, side by side.
Ask for access
More from the playground
All tests →
Export
Include

Chrome will ask to share this tab: choose Allow. Only the board below is recorded.

Spring Promptspringprompt.com