What's New — daily benchmark digest
New benchmark leaders
- Kimi K3.1 Pro took the lead on Humanity's Last Exam (Self-Reported, No Tools) (51.1 vs 44.6 by Ornith-1.5-397B)
- Kimi K3.1 Pro took the lead on Humanity's Last Exam (Self-Reported, With Tools) (70 vs 63.9 by DeepSeek V4.1 Flash)
- Bugstack + Claude Sonnet 5.5 + Opus 5.5 took the lead on SWT-Bench Verified (91.5 vs 87.5 by DPIAgent + GPT-5 (with test selection))
- Keen AgenticOS V1.1 took the lead on GAIA (94.35 vs 94.02 by Shadow Harness V8.01)
New benchmarks
New models
Interactive version: theaggregate.ai/whats-new · How It Works · Data refreshed daily, snapshot 2026-11-05.