Added Grok 4.6
xAI's Grok 4.6 ran the full matrix clean — 35 of 35 builds shipped — and averaged 3.43, with a clean five and a clear strength in generative canvas work.
xAI shipped Grok 4.6, so we added it to the matrix and ran all 35 one-shot prompts against it — same harness, same single-shot rules as every other build in the gallery.
It ran clean: 35 of 35 builds shipped, averaging 3.43, with 19 builds at four stars or better and one clean five — a flow-field particle system with evolving color and reactive input. It's at its best on generative canvas work and physics: a fully simulated aquarium that cracks, jets, and drains onto the floor; smooth cursor-fleeing boids; a dye-based fluid simulation; a colorful double pendulum.
Its misses cluster in one place — game-start logic. Several games render a polished menu and then never begin: a frozen Mario scene with no response to input, a Snake stuck on "Ready," a Tetris that won't leave its title screen. Nothing blank or crashed, just play loops that don't quite ignite.
The top builds below are its best-graded outputs; open any one for the live artifact, the full transcript, and the two-tier evaluation behind the score.
Top builds in this update
CLEAN “Flow-field particle animation with fading trails, evolving colors, and reactive click/drag/wheel stirring effects; arrows/wasd no effect (not needed).”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
CLEAN “Aquarium with fish, plants, crack, burst jet, draining waterline, and spreading puddle all render correctly on space trigger.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
CLEAN “Boids flock smoothly with trailing triangles and flee from cursor position shown as red glow; sliders present but idle otherwise.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
CLEAN “Neon-styled breakout with menu, paddle, ball, bricks breaking, score, lives; click, arrows, wasd all affect gameplay.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →