Language modelActive

Hermes 4 405B

Nous Research

Released: August 6, 2025
Data date: September 8, 2026
Predecessor: Hermes 3 405B (Official documentation)

Speed: 41.6 tokens/s
Output throughput
Price: $1 · $3
Input · Output / 1M tokens
Input: Text
Output: Text

Hermes 4 405B combines a visible reasoning mode with direct answers and structured outputs. The open dense Llama 3.1 model has 405 billion parameters, a 131,072-token context, and tool-call support.

Nous Research describes Hermes 4 as an update to the 405B tier of Hermes 3 and lists five million post-training examples plus 56 billion training tokens. The weights use the Llama 3.1 Community License, not MIT or Apache 2.0.

Specifications and access

Specification	Value and source
Weights	Open weightsSource
Access	API, LocalSource
Parameters	405BSource
Context window	131,072 tokensSource
License	Llama 3.1 Community LicenseSource
Parameters	405 BSource
Architecture	dense Llama 3.1 decoder-only transformer, 126 layersSource
Input	TextSource
Output	TextSource
Reasoning mode	hybrid mode; model can emit explicit <think> reasoning or answer in non-reasoning modeSource
Tool calling	YesSource
Structured output	YesSource
Post-training samples	5,000,000Source
Post-training tokens	60 B (The technical report's 56B figure is retained separately for the reported training setup.)Source
Training tokens	56 BSource
Training context	16,384 tokensSource

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

MATH500 (reasoning) (Show measurement, test conditions, and source)

Source value: 96.3
Score: 96.3
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

MATH500 (non-reasoning) (Show measurement, test conditions, and source)

Source value: 73.8
Score: 73.8
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

AIME24 (reasoning) (Show measurement, test conditions, and source)

Source value: 81.9
Score: 81.9
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

AIME24 (non-reasoning) (Show measurement, test conditions, and source)

Source value: 11.4
Score: 11.4
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

AIME25 (reasoning) (Show measurement, test conditions, and source)

Source value: 78.1
Score: 78.1
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

AIME25 (non-reasoning) (Show measurement, test conditions, and source)

Source value: 10.6
Score: 10.6
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

GPQA (reasoning) (Show measurement, test conditions, and source)

Source value: 70.5
Score: 70.5
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

GPQA (non-reasoning) (Show measurement, test conditions, and source)

Source value: 39.4
Score: 39.4
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

BBH (reasoning) (Show measurement, test conditions, and source)

Source value: 86.3
Score: 86.3
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

BBH (non-reasoning) (Show measurement, test conditions, and source)

Source value: 68.7
Score: 68.7
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

LCBv6 (reasoning) (Show measurement, test conditions, and source)

Source value: 61.3
Score: 61.3
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

LCBv6 (non-reasoning) (Show measurement, test conditions, and source)

Source value: 28.1
Score: 28.1
Metric: percent
Unit: %
Category: mixed
Direction: Higher is better
Source type: vendor-reported
Evaluator: Hermes 4 Technical Report Table 3
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Official report Table 3; each row preserves the R (reasoning) or N (non-reasoning) column.

Source Hermes 4 Technical Report Table 3

Page 1 of 4

Predecessor and successor

Sources and data date

Type	Evidence and data date
Speed	Median across serving providers, Non-reasoning. Output speed after the first streaming chunk, not total response time. Retrieved September 8, 2026. Source: Artificial Analysis · Artificial Analysis (retrieved September 8, 2026)Speed measurement methodology
Price	Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. Source: Pricing · API pricing
Input	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Nous Research Hermes 4 405B model card · Editorial description reviewed October 2, 2026
Output	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Nous Research Hermes 4 405B model card · Editorial description reviewed October 2, 2026
Research status	Research date September 8, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	active-parameters-billions Original wording The exact model is dense 405B; no separate active-parameter count applies or is published.
Not found in the inspected sources	knowledge-cutoff Original wording The exact card/report does not publish a knowledge cutoff.
Additional source	Hermes 4 Technical Report Table 3 (retrieved September 8, 2026) · Nous Research Hermes 4 technical report · Editorial description reviewed October 2, 2026
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Creative Commons Attribution 4.0

Hermes 4 405B

Specifications and access

Measurements without matching peer values

Our tests

Pelican on a bicycle

Popularity

Published prices

Predecessor and successor

Sources and data date