Nous ResearchOpen weightsReasoning modelAuto-detected

Hermes 4 - Llama-3.1 405B

Name: Hermes 4 - Llama-3.1 405B benchmark results
Creator: BenchLeader

Best configuration ranks #423 of 610 on the BenchLeader Index at 44.1 ±3.8. Released 27 Aug 2025.

Blended price: $1.50/M; $1.00 in · $3.00 out
Output speed: 43 tok/s
First answer: 2.40 s; first token 2.40 s
Context: 128k

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index44
Agents & tools42
Knowledge48
Instruction following38
Long context36
Composite38

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
thinking	43.9	#430	43 tok/s	49 s	$0.0013	Agents & tools 43 · Composite 38 · Instruction following 36 · Knowledge 47 · Long context 36
defaultbest	44.1	#423	43 tok/s	2.40 s	$0.0013	Agents & tools 42 · Composite 38 · Instruction following 38 · Knowledge 48 · Long context 36

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	thinking	default	Source
GPQA Diamond (AA)not in index	72.7%#264	53.6%#390	Artificial Analysis
Humanity's Last Exam (AA)not in index	10.9%#242	4.2%#441	Artificial Analysis

Agents & tools

Benchmark	thinking	default	Source
Terminal-Bench Hard	11.4%#207	9.8%#220	Artificial Analysis
τ²-Bench Telecom (AA)not in index	22.2%#300	26.6%#271	Artificial Analysis

Knowledge

Benchmark	thinking	default	Source
AA-Omniscience	-36.0#237	-33.3#228	Artificial Analysis

Instruction following

Benchmark	thinking	default	Source
IFBench	32.7%#323	34.8%#298	Artificial Analysis

Long context

Benchmark	thinking	default	Source
AA-LCR	22.3%#350	22.0%#352	Artificial Analysis

Composite

Benchmark	thinking	default	Source
AA Intelligence Index	7.5#392	7.4#397	Artificial Analysis

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0013	–	9.3 s
Summarise a 30-page report	12,000 / 600	$0.014	–	16.3 s
Code edit	6,000 / 1,500	$0.011	–	37.0 s
Agentic coding session	60,000 / 4,000	$0.072	–	1.6 min
Structured extraction	2,000 / 200	$0.0026	–	7.0 s