K2 Horizon 375B A23B: benchmark results
Provider: LLM360. Released 2026-09-03. Access: Open.
Unified ELO 1753 ± 33, rank #114 of 1645 rated models, from 67 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| IFM K2-Horizon-375B Card - SWE-Atlas QnA (Strict, No Internet) | 48.4 | Score (%) | 100 |
| AA-Omniscience Hallucination Rate | 25.85 | Hallucination Rate (%) | 95.3 |
| AA Long Context Reasoning | 80 | Accuracy (%) | 86.6 |
| AA-Omniscience Index - Health | -2.9 | Omniscience Index | 84.2 |
| AA GPQA Diamond | 87.27 | Accuracy (%) | 83.7 |
| Artificial Analysis Intelligence Index | 30.5 | Intelligence Index | 83.6 |
| AA-Omniscience Index - Science, Engineering & Mathematics | 7.8 | Omniscience Index | 82.7 |
| AA Humanity's Last Exam | 31.97 | Accuracy (%) | 79.2 |
| AA-Briefcase - DOCX - Rubric Pass Rate | 47.2 | Rubric Pass Rate (%) | 78.9 |
| AA-Omniscience Index - Business | -3.8 | Omniscience Index | 77.2 |
| AA-Omniscience Index - Software Engineering (SWE) - Julia | 4 | Omniscience Index | 76.1 |
| CritPt | 4.6 | Accuracy (self-reported) | 75.3 |
Interactive version: theaggregate.ai/model?slug=k2-horizon-375b-a23b · How It Works · Data refreshed daily, snapshot 2026-10-10.