DeepHermes 3 - Mistral 24B Preview (Non-reasoning) — benchmark results

Provider: Nous Research. Released 2025-01-15. Access: Open.

Unified ELO 1403 ± 48, rank #1257 of 1841 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Epoch AI - Scicode22.8Score23.3
AA MMLU-Pro58.01Accuracy (%)21.5
AA LiveCodeBench19.47Pass@1 (%)20.6
AA MATH-50059.47Accuracy (%)20.4
Artificial Analysis Intelligence Index5.31Intelligence Index16.4
AA GPQA Diamond38.18Accuracy (%)15.2
AA Humanity's Last Exam3.92Accuracy (%)10.7

Interactive version: theaggregate.ai/model?slug=deephermes-3-mistral-24b-preview-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-07-25.