Chiikabench - Full Cohort (250): leaderboard
Chiikabench smart-contract challenge suite spanning five chains. Recon/orchestrator/builder/verifier agent team; solved when extraction event meets per-challenge capture floor. The source publishes no token or effort budget. Oracle is a reference, not an evaluated model. DeepSeek reported attempted-subset denominator excludes 23 rejections and remains separate from the full250challenge cohort. Full Cohort (250). Metric: Pass rate (%).
Metric: Pass rate (%). Source: huggingface.co. Saturation forecast: Around 2031. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.3 Flash | 2.8 | #76 |
| 2 | Kimi K3 | 2.4 | #34 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=chiikabench-full-cohort-250 · How It Works · Data refreshed daily, snapshot 2026-10-27.