Qwen 3.8 27B NVFP4 Think: medium | Benchmark #156

Described in:Counter-Strike 2D-like game built by Qwen 3.6 and Qwen 3.8 as agents

Benchmark summary

Counter-Strike 2D-like game built in SC# as an agent

ModelQwen 3.8 27B QuantizationNVFP4 ModeThink: medium

The model builds a Counter-Strike 2D-like top-down shooter 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.

Score 18/18 100.0% passed
18 passed18 total 1 AI query 21 tool calls TOKENS IN1 035 652 TOKENS OUT98 172

Results by difficulty

1 group
Level Unknown / other
18 / 18 100.0%
passed
1 AI query 21 tool calls IN1 035 652 OUT98 172

Results by category

1 group
Category Other
100.0% 18/18 passed
1 AI query 21 tool calls TOKENS IN1 035 652 TOKENS OUT98 172

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#: Counter-Strike 2D-like top-down shooter 18 passed
100.0%
IN1 035 652 OUT98 172
✓
TopDownShooterThe model builds a top-down shooter in the style of Counter-Strike 2D 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 21 tool calls
Passed IN1 035 652 OUT98 172
✓
Hidden_ArmorA bot bullet does 10 damage, and armor takes half of it.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_BotsPatrolBots patrol: they keep moving, at most 1.5 px per frame, and never stand in a blocked position.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_BotsShootA bot that sees the player shoots at him after a reaction time of 10 to 30 frames, then every 20 to 40 frames, aiming at his center.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_BulletsStopWalls and crates stop bullets.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_FiringSpace fires a pistol bullet from 8 px in front of the player at 8 px per frame; the next shot comes 8 frames later; with no ammo nothing is fired.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_HittingBotsA pistol bullet takes 20 health from a bot; a bot without health is removed, and without bots the round is won.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_InitialStateAfter AppInit the map, the player, the bots, the pickups and the counters have their starting values.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_LosingARoundWhen the player's health drops to 0 the round is lost; the next round starts without armor and with the pistol.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_MovingThe player moves 2 px per frame in 8 directions, diagonals included.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_NoLineOfSightA bot never shoots without line of sight: a wall between them blocks it.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_PickupsPickups are taken only when they help: health +50 up to 100, armor to 100.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_ReloadingR reloads the pistol in 45 frames; during the reload the player cannot fire; R does nothing with a full magazine.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_RifleThe rifle fires every 3 frames, does 30 damage and reloads in 75 frames.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_ScreenShowsTheMapThe screen shows the map tiles, the pickups, the bots, the player and the texts.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_ShopIn the shop, 1 buys armor for 200 and 2 the rifle for 400; after a won round armor and weapon stay.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_TurningZ and X turn the aim by 6 degrees per frame, kept between 0 and 360.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0
✓
Hidden_WallsAndSlidingWalls stop the player; moving at a slant into a wall slides along it.0 AI queries 0 tool callsHiddenTest
Passed IN0 OUT0