Gallery, prompt v2
A serving tray where every table gets a drink it will take
One frozen prompt and every model that has run it, newest first.
Saturated since 2026-11-08
It is off the board. Every run of all eight models that drew a tray served all six tables a drink they will take. Three runs ran out of time with nothing drawn: Kimi K3.1 Pro, GPT-6 Luna and GPT-6.1 Sol, one each. A six by six matching is too small to tell models apart, and the harder version is nine tables and nine drinks.
The prompt, v2, frozen
I'm a barista and I have six drinks ready to go out: - Vanilla cortado: dairy milk, vanilla syrup, hot - Decaf cortado: dairy milk, decaf, hot - Decaf oat vanilla latte: oat milk, decaf, vanilla syrup, hot - Iced oat latte: oat milk, cold - Cold brew: black, cold - Filter coffee: black, hot There are six tables, and each one has asked for something: - Table 1 only drinks dairy milk - Table 2 wants decaf - Table 3 wants vanilla - Table 4 wants oat milk - Table 5 wants something cold - Table 6 wants black coffee Each table gets one drink. Each drink goes to one table. Draw the tray as tray.svg. Label every cup with its table number and the drink's name.
Every model, its first three runs.
One square a check, in this order:
- Serves all six
passed, failed, undecided. A round mark is a check I judged by eye, and a diamond one a check a model judged.
| Kimi K3.1 Pro | |||
| GPT-6.1 Sol | |||
| GPT-6 Luna | |||
| Opus 5.5 | |||
| Claude Sonnet 5.5 | |||
| Claude Haiku 5.5 | |||
| DeepSeek 4.1 Flash | |||
| GLM 5.3 FlashX |
How it is checked
Six drinks with fixed attributes, six tables with fixed notes, one tray. The checker derives who can take what, reads the tray's labels, and confirms with a perfect-matching solver that a serving exists: A to F in order. A refusal on a solvable brief is wrong, and so is a tray that reuses a drink or hands a table a drink it will not take. Whether a drawn cup really holds what its label says is the eye's.
What replaces it: Nine tables and nine drinks, so the matching is not read off six names.
The rules behind all of thisThis task's prompt, checks and any checker, on GitHub
Earlier results ran under an older prompt, so they are at the bottom, next to the question they were asked.
Older prompts
Prompt v1, retired 2026-11-08
Six drinks are ready on the pass, each with a card saying what is in it: - Vanilla cortado: dairy milk, regular, vanilla syrup, hot - Decaf cortado: dairy milk, decaf, no syrup, hot - Decaf oat vanilla latte: oat milk, decaf, vanilla syrup, hot - Iced oat latte: oat milk, regular, no syrup, cold - Cold brew: black, regular, no syrup, cold - Filter coffee: black, regular, no syrup, hot Six tables are waiting, each with a note of what it will take: - Table 1: dairy milk only - Table 2: no caffeine - Table 3: vanilla - Table 4: oat milk - Table 5: something cold - Table 6: black coffee Draw the serving tray as tray.svg: six cups, each labelled with the table number and the drink's name, so every table gets a drink it will take, every drink is used exactly once, and no drink goes to two tables.
The first wording read like a model wrote it: drinks "on the pass" with cards, tables with notes, and a last line that said the same rule three times. The facts did not change.
Nothing run on this prompt yet.