qwen3-coder:30b vs qwen3:14b

Both models answered the same one-shot tasks under identical rules — no retries, answers published unedited. Two separate axes, never blended: does the code actually run, and how do blind-scored text answers hold up against a written gold answer.

Model	Runnable sims	Blind text avg	Failure modes
qwen3-coder:30b	24% (23/95)	9.17 (48 scored)	ERR 41 · STAT 9 · BLANK 21 · HANG 1
qwen3:14b	31% (29/95)	8.73 (48 scored)	ERR 42 · STAT 10 · BLANK 13 · HANG 1

Biggest contrasts

Rounds where the two models diverged most — open them and read both answers side by side.

Round	qwen3-coder:30b	qwen3:14b
R175 · Noise flowfield painting additive particle trails	BLANK	OK
R174 · Recursive fractal tree growing branch by branch	BLANK	OK
R172 · Raindrops sliding down a window pane	ERR	OK
R163 · Prim's algorithm growing a minimum spanning tree	ERR	OK
R162 · Breadth-first search flooding a random grid maze	OK	ERR

→ Interactive round-by-round comparison · → Full leaderboard

Generated 2026-08-04 from the published rounds. Harness failures are excluded from every base — a broken cluster node is our fault, not the model's.