Head-to-head
mistral-small:24b vs GPT-5.5 (Codex)
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 |
|---|---|---|---|
| mistral-small:24b | 32% (30/95) | 8.26 (38 scored) | ERR 29 · STAT 12 · BLANK 9 · HANG 7 · DNF 8 |
| GPT-5.5 (Codex) frontier reference | 96% (91/95) | unscored by design | ERR 2 · STAT 1 · BLANK 1 |
Biggest contrasts
Rounds where the two models diverged most — open them and read both answers side by side.
| Round | mistral-small:24b | GPT-5.5 (Codex) |
|---|---|---|
| R175 · Noise flowfield painting additive particle trails | STAT | OK |
| R174 · Recursive fractal tree growing branch by branch | HANG | OK |
| R173 · Metaball lava lamp with heating and cooling blobs | ERR | OK |
| R171 · Fireworks bursting into fading sparks | HANG | OK |
| R165 · Recursive backtracker carving a maze, then solving it | HANG | OK |
→ 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.