Benchmark summary
The model builds a Diamond Dash-like gem clicker (groups of 3 or more gems vanish, the rest fall) played with the mouse with a polished look in SC# as an agent: it writes the app, compiles it, runs its own tests and fixes the code. Hidden tests check the rules from the task; each hidden test is its own result.
Tests with multiple category tags are counted in each matching category, so category totals may exceed the report total.
| Benchmark items | |||
|---|---|---|---|
| Script / test | Status | Tokens | |
Building Games in SC#: Diamond Dash-like gem clicker
15 passed |
100.0% |
IN1 061 436 OUT68 220 | |
| ✓ | DiamondDashThe model builds a mouse game like Diamond Dash (click groups of 3 or more matching diamonds, they fall and refill, a 60-second round with combos) from the task as an agent; this item passes when the app compiles and can be exported, and each hidden test is its own item.1 AI query
20 tool calls |
Passed | IN1 061 436 OUT68 220 |
| ✓ | Hidden_ClicksOutsideAndAtTheEdgesClicks outside the board do nothing; the board edges belong to the edge cells.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_ComboA hit within the combo time raises the combo (at most 5) and multiplies the score; after the time runs out the combo starts again at 1.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_EffectsEndThe effects of a hit and of a miss end within a second: afterwards the still board looks exactly as before the miss.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_FallAndRefillOrderThe diamonds above a removed group keep their order, and columns are refilled from left to right, each from its lowest empty cell up.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_GroupsAreOrthogonalGroups connect only through left, right, up and down neighbors, and the whole group is removed however it winds.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_HitRemovesAGroupA click on a group of 3 removes it, the diamonds above fall down and the new ones come from refill, bottom first.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_HoverHighlightThe mouse over a group of 3 or more highlights exactly the cells of that group; over a smaller group nothing changes on the board.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_InitialStateAfter AppInit a full board of colors 1 to 5 is ready, the refill list is empty and the counters have their starting values.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_MissLocksClicksA click on a group of 1 or 2 is a miss: nothing is removed, the combo ends and clicks are ignored for the next 30 frames.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_NewGameEnter starts a new game at any time, also after the time is up; a click in the same frame does nothing.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_OwnRefillWith an empty refill list the game picks the new colors itself, always from 1 to 5.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_ScoreBySizeThe score of a hit is 10 * n * (n - 2) for 3, 4 and a whole column of 9 diamonds.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_ScreenShowsTheDiamondsWhen the board is still, the middle pixel of every cell shows its diamond's color: same color, same pixel; different colors, different pixels; also after a hit.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |
| ✓ | Hidden_TimeRunsOutThe time counts down once per frame and is shown in whole seconds rounded up; at 0 the round ends and clicks do nothing.0 AI queries
0 tool callsHiddenTest |
Passed | IN0 OUT0 |