Models/Hermes 4 - Llama-3.1 405B

MDL

Hermes 4 - Llama-3.1 405B

Nous Research

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$1 / $3

in / out

Context

—

Avg score

41.6

10 benchmarks

AA Intelligence

9.0

index

Speed

41 tok/s

TTFT 0.7s

Released

Aug 27, 2025

About

Hermes 4 - Llama-3.1 405B is a open-weights model in the Hermes family from Nous Research. Benchmarked on 10 evals, averaging 41.6.

Benchmark scores · 10

avg 41.6 — higher is better

MMLU-Proartificial-analysis · 2026-07

82.9

GPQA Diamondartificial-analysis · 2026-07

72.7

AIMEartificial-analysis · 2026-07

69.7

LiveCodeBenchartificial-analysis · 2026-07

68.6

IFBenchartificial-analysis · 2026-07

32.7

SciCodeartificial-analysis · 2026-07

25.2

τ²-benchartificial-analysis · 2026-07

22.2

AA Long-Context Reasoningartificial-analysis · 2026-07

20.7

Terminal-Bench Hardartificial-analysis · 2026-07

11.4

Humanity's Last Examartificial-analysis · 2026-07

10.3

Details

PublisherNous Research

FamilyHermes

Parameters—

License—

WeightsOpen

ReleasedAug 27, 2025

Performance & price

Output speed41 tok/s

Latency (TTFT)0.7s

First answer49s

Price in / out$1 / $3 per 1M

Blended price$1.5 / 1M

Math index69.7

Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$3/M