Overview
Independent AI benchmarking platform frequently cited for open- vs closed-model Intelligence Index and related metrics (e.g. Omniscience factual accuracy).
Recent Developments
-
2026-07: Ranks hosted kimi-k3 (~4th per secondary digests); classifies as proprietary until open weights
-
2026-07-16: Cited for inkling Intelligence Index ~41 (top U.S. open-weights) and weak Omniscience / high hallucination stats (2026-07-16-thinking-machines-inkling-decoder)