Hermes 4 70B FP8

Name: Hermes 4 70B FP8
Author: Nous Research

by Nous ResearchMeta

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released August 2025131K context≈ 98,304 words70B params

Hermes 4 70B FP8 is NousResearch's fine-tune of Meta's Llama 3 architecture, known for strong instruction-following and structured output capabilities. The FP8 quantization keeps the 70B parameter footprint manageable without dramatically sacrificing quality, making it practical for self-hosted deployments. It handles long contexts up to 128K tokens and tends to be reliable for agentic and tool-use workflows.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Long Context

Moderate

Multilingual

Hermes 4 70B FP8

by Nous ResearchMeta

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released August 2025131K context≈ 98,304 words70B params

Hermes 4 70B FP8 is NousResearch's fine-tune of Meta's Llama 3 architecture, known for strong instruction-following and structured output capabilities. The FP8 quantization keeps the 70B parameter footprint manageable without dramatically sacrificing quality, making it practical for self-hosted deployments. It handles long contexts up to 128K tokens and tends to be reliable for agentic and tool-use workflows.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Long Context

Moderate

Multilingual

Glossary

AgenticA model designed to act autonomously by making decisions, selecting actions, and using tools to accomplish multi-step tasks.ArchitectureThe underlying structural design of a neural network that defines how data flows through layers and components.FP8 QuantizationA compression technique that reduces model size by representing weights using 8-bit floating-point numbers instead of higher precision, making it faster and more memory-efficient.Fine-TuneA model created by training an existing pre-trained model on new data to specialize it for specific tasks or behaviors.Instruction-FollowingThe ability of a model to understand and execute specific tasks or commands given in natural language prompts.Parameter FootprintThe total number of learnable weights in a model, which directly affects its memory requirements and computational cost — smaller footprints run faster on consumer devices.QuantizationReducing a model's numerical precision (e.g., from 16-bit to 4-bit) to shrink memory usage and speed up inference.Self-HostedRunning a model on your own hardware and infrastructure instead of relying on a company's servers or API.Structured OutputResponses formatted in a consistent, machine-readable way (like JSON or XML) rather than free-form text.TokensThe basic units of text that a language model processes, typically representing words or word fragments.

Capabilities

Capabilities

Use Case Fit

Glossary