Nano GPT logo
NanoGPT
Provider logo

Hermes 4 Large (Thinking)

nousresearch/hermes-4-405b:thinking
BackTry Model
Provider logo

Hermes 4 Large (Thinking)

nousresearch/hermes-4-405b:thinking
BackTry Model

Hermes 4 Large with thinking enabled. Streams visible reasoning before the final answer and supports structured outputs.

Model weights

Context Window

128.0K

Max Output

8.2K

Avg output tokens (7d)

1.1K tokens

47%

Input Price (Auto)

$0.30/1M

Output Price (Auto)

$1.20/1M

Cache Read (Auto)

$0.15/1M

Capabilities

Benchmarks

Performance metrics and benchmarks

Sourced from Artificial Analysis.

Intelligence Index

8.8

Better than 27% of models compared

Reasoning

GPQA Diamond

Graduate-level scientific reasoning

72.7%

Better than 55% of models compared

HLE

Humanity's Last Exam

10.9%

Better than 60% of models compared

IFBench

Instruction-following benchmark

32.7%

Better than 21% of models compared

T²-Bench Telecom

Conversational AI agents in dual-control scenarios

22.2%

Better than 24% of models compared

AA-LCR

Long context reasoning evaluation

22.7%

Better than 31% of models compared

CritPt

Research-level physics reasoning

0.3%

Coding

SciCode

Python programming for scientific computing

25.2%

Better than 29% of models compared

Terminal-Bench Hard

Agentic coding and terminal use

11.4%

Better than 47% of models compared

LiveCodeBench

Contamination-free coding benchmark

68.6%

Better than 77% of models compared

Math

AIME 2025

American Invitational Mathematics Examination 2025

69.7%

Better than 64% of models compared

Knowledge

MMLU-Pro

Professional and academic subject knowledge

82.9%

Better than 82% of models compared

AA-Omniscience Accuracy

Proportion of correctly answered questions

30.1%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

94.5%

Last updated Aug 16, 2026

Artificial Analysis

Providers

Auto routing is available for this model. Explicit provider selection is not available.

Loading provider options…

Compare Hermes 4 Large (Thinking) with similar models from the same provider or model family.

Hermes 4 Large

nousresearch/hermes-4-405b

Advanced reasoning model built on Llama-3.1-405B with hybrid thinking modes. Features internal deliberation capabilities, excels at math, code, STEM, and logical reasoning while supporting structured outputs with improved steerability and neutral alignment.

Hermes 3 70B

nousresearch/hermes-3-llama-3.1-70b

Hermes 3 is a generalist language model with many improvements over Hermes 2, including advanced agentic capabilities, better roleplaying, reasoning, multi-turn conversation, and long context coherence. This 70B model is a competitive finetune of Llama-3.1-70B focused on aligning LLMs to the user with powerful steering capabilities.

Hermes 4 (Thinking)

NousResearch/Hermes-4-70B:thinking

Hermes 4 70B with thinking enabled. Emits explicit reasoning content before final answer when streamed.

Hermes 4 Medium

nousresearch/hermes-4-70b

Efficient reasoning model based on Llama-3.1-70B. Offers hybrid thinking capabilities with strong performance in math, code, and logical reasoning tasks. Supports structured outputs and JSON mode with enhanced steerability.

Hermes High

hermes-high

Currently points to Claude 4.7 Opus. Highest tier for Hermes-style agent work: most expensive, strongest intelligence, and best suited for hard multi-step tasks.

Hermes Low

hermes-low

Currently points to Gemini 3.1 Flash Lite. Lowest tier for Hermes-style agent work: lowest cost for routine, fast, high-volume tasks.