OverviewCompare › gemma4:26b vs GPT-5.5 (Codex)
Head-to-head

gemma4:26b 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.

ModelRunnable simsBlind text avgFailure modes
gemma4:26b 0% (0/1)9.61 (31 scored)BLANK 1
GPT-5.5 (Codex) frontier reference 95% (81/85)unscored by designERR 2 · STAT 1 · BLANK 1

Biggest contrasts

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

Roundgemma4:26bGPT-5.5 (Codex)
R117 · De Casteljau construction of a Bezier curveBLANKOK

→ 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.