← Evaluations/EVAL-20260402-145527
code
Apr 02, 2026CODE-027

Implement a JSON Schema validator from scratch that supports: type validation (string, number, integer, boolean, array, object, null), required fields, minimum/maximum for numbers, minLength/maxLength for strings, pattern (regex), enum, nested objects, and array items validation. No external libraries.

Winner
GPT-5.4
openrouter
8.96
WINNER SCORE
matrix avg: 6.77
results.json report.mdFull dataset (CSV) →
10×10 Judgment Matrix · 90 judgments
OPEN DATA
Judge ↓ / Respondent →GPT-OSS-120BGPT-5.4Claude Opus 4.6Gemini 3.1 ProClaude Sonnet 4.6Grok 4.20DeepSeek V4Gemini 3MiniMax M2.5MiMo-V2-Flash
GPT-OSS-120B8.83.82.14.48.89.08.83.48.3
GPT-5.44.42.50.73.57.88.08.31.45.2
Claude Opus 4.67.49.01.65.37.88.68.01.97.0
Gemini 3.1 Pro5.29.84.56.48.29.47.83.26.3
Claude Sonnet 4.67.88.66.81.98.88.67.84.37.4
Grok 4.208.68.68.50.85.87.37.05.56.8
DeepSeek V49.48.68.88.28.48.68.87.08.6
Gemini 39.29.86.32.37.89.89.85.59.6
MiniMax M2.57.48.85.00.46.78.88.68.67.0
MiMo-V2-Flash8.68.64.48.37.79.08.88.60.2