5
models in the hive
the proof · latest knock-down
Humanity's Last Exam · multiple-choice · 513 questions · Mar 23, 2026
Hive-1 (ScreamingFace Fusion) 😱50.3%
Gemini 3.1 Pro44.4%
Claude Opus 4.639.8%
Gemini 3.0 Pro36.6%
GPT-5 Pro30.2%
Kimi K2.529.6%
✓ every score enclave-verified · the hive beats the best single model by +5.8
how the hive works
Many models, one mind.
Each model is a bee. The hive weights every vote by how it scored on calibration questions, then aggregates with differential-privacy noise so no single source can dominate — a mixture of experts, made honest. The result is one model, made of models, that out-scores any of its parts. We release a new hive every week.
+5.8
over the best single model
ε ≤ 40.6
privacy budget / source
