Benchmark Almanac

Hermes 4 405B

Hermes 4 405B places in the top three on 2 benchmarks.

Nous Research Released August 26th, 2025 Family Hermes 4 The lab's model card Compare with others
21
Benchmarks with a score
0
Ranked first, out of five or more models
2
In the top three
Nous Research Aug 2025

Hermes 4 405B

Open weights Flagship

135.0 #230 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • MMLU-Pro 80.6% #75 /167
  • LiveCodeBench v6 61.4% #74 /90
  • AIME 2025 78.1% #91 /143

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
Arena-Hard-Auto 94.4% #3 of 39 Lab model card
OpenBookQA 94.2% #2 of 24 Lab model card
BIG-Bench Hard 86.3% #21 of 73 Lab model card
MMLU 87.2% #39 of 131 Lab model card
EQ-Bench 3 1,405 #33 of 78 Official leaderboard
DROP 83.5% #18 of 41 Lab model card
MMLU-Pro 80.6% #75 of 167 Lab model card
SimpleQA 25.8% #29 of 60 Lab model card
Creative Writing v3 66.7 #63 of 120 Official leaderboard
MuSR 66.1% #4 of 6 Lab model card
Longform Creative Writing (EQ-Bench) 51.2 #81 of 119 Official leaderboard
IFEval 81.5% #101 of 135 Lab model card
RewardBench 73% #2 of 2 Lab model card

Math and science

BenchmarkScoreRankSource
AIME 2024 81.9% #35 of 88 Lab model card
MATH-500 96.3% #30 of 70 Lab model card
GPQA Diamond 70.6% #181 of 302 Lab model card
AIME 2025 78.1% #91 of 143 Lab model card

Coding

BenchmarkScoreRankSource
LiveCodeBench v6 61.4% #74 of 90 Lab model card

Agents and tools

BenchmarkScoreRankSource
SnakeBench 21 #128 of 188 Official leaderboard

Safety and security

BenchmarkScoreRankSource
nocot-bench (NCRI) 62.8 #130 of 185 Official leaderboard
RefusalBench (Nous Research) 57.1% #2 of 2 Lab model card

Its lab's numbers put it at 135.0, number 230 of 404

The Self-Reported Index fits one score to the 13 capability results Nous Research reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 132.3 to 137.7. How we compute it.

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
Creative Writing v3 79.8 CreativeWriting3 (reasoning) 66.7 Leaderboard +13.1

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Hermes 4 405B model card (Hugging Face) Aug 26, 2025

Its system card

Every result in the longer documents the lab published with it.