Hermes 4 Technical Report

TL;DR

Hermes 4 introduces a family of hybrid reasoning models that integrate structured multi-turn reasoning with broad instruction-following, demonstrating strong performance across diverse benchmarks and supporting open research with publicly available weights.

Contribution

The paper presents Hermes 4, a novel hybrid reasoning model family combining structured reasoning with instruction-following, addressing data and evaluation challenges at scale.

Findings

Strong performance on mathematical reasoning benchmarks

Effective coding and knowledge understanding capabilities

Positive qualitative behavioral analysis

Abstract

We present Hermes 4, a family of hybrid reasoning models that combine structured, multi-turn reasoning with broad instruction-following ability. We describe the challenges encountered during data curation, synthesis, training, and evaluation, and outline the solutions employed to address these challenges at scale. We comprehensively evaluate across mathematical reasoning, coding, knowledge, comprehension, and alignment benchmarks, and we report both quantitative performance and qualitative behavioral analysis. To support open research, all model weights are published publicly at https://huggingface.co/collections/NousResearch/hermes-4-collection-68a731bfd452e20816725728