Le Chonk on trial: Mistral Large 4 vs Opus 5.5 vs GPT-6.1 Sol

Cherry trees, torii gate, lanterns, clouds under the island; a warm sunset, but the night is almost black and the windows do not light up. Reasoning low (on high it thought for 100k tokens and returned no file). No edits.

From our Oct 6 test: koi pond, falling petals, a waterfall; at night lanterns glow and fireflies come out. No edits.

From our Oct 6 test: detailed, with a starry night, but the camera starts close. No edits.

Card in, Processing, a green tick, a receipt; the screen texts overlap. Reasoning off (on high and on low it spent its whole output thinking and returned no file). No edits.

From our Oct 6 test. No edits.

From our Oct 6 test: a keypad that types a PIN, a receipt with line items. No edits.

Thirty parts registered as clickable, but walls and beams sit in the wrong places and the appliances are hidden behind them, so clicks do nothing. Reasoning off (on low it ran out of output while thinking; a first reasoning-off try stopped after one sentence with a tool call). No edits.

From our Oct 6 test: 32 parts you can click, daylight and pendant lights. No edits.

From our Oct 6 test: a dollhouse with 17 working parts. One edit (a duplicate name, marked in the source).

Title screen, countdown, minimap and speedometer, but the track tilts on its side, the lap counter never moves, and Space boost cannot work (the code waits for a key named "space"). Reasoning off (two runs with reasoning on returned no file). No edits.

From our Oct 6 test: banked turns, a tunnel of rings, five rivals. No edits.

From our Oct 6 test: fast, with drifting and boost pads. No edits.