Head-to-head
qwen3.5:9b vs Claude Sonnet 5
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. Frontier references sit outside the blind field and are deliberately unscored on text.
| Model | Runnable sims | Blind text avg | Failure modes |
|---|---|---|---|
| qwen3.5:9b | 36% (31/85) | 9.21 (38 scored) | ERR 26 · STAT 11 · BLANK 16 · HANG 1 |
| Claude Sonnet 5 frontier reference | 100% (85/85) | unscored by design | — |
Biggest contrasts
Rounds where the two models diverged most — open them and read both answers side by side.
| Round | qwen3.5:9b | Claude Sonnet 5 |
|---|---|---|
| R155 · Linked list with animated pointer relinking | STAT | OK |
| R152 · Hash table with visible collision chains | STAT | OK |
| R151 · Binary search tree growing node by node | ERR | OK |
| R139 · Small n-body system orbiting a central star | STAT | OK |
| R137 · Cloth mesh hanging and rippling in the wind | ERR | OK |
→ Interactive round-by-round comparison · → Full leaderboard
Generated 2026-07-31 from the published rounds. Harness failures are excluded from every base — a broken cluster node is our fault, not the model's.