Unveiling Hermes 3: The First Full-Parameter Fine-Tuned Llama 3.1 405B Model is on Lambda’s Cloud
Blog post from Lambda
Hermes 3`, the first full-parameter fine-tuned `Llama 3.1 405B model`, is now available on `Lambda's Cloud`. This neutral generalist model, designed for the open-source community, offers exceptional reasoning capabilities and meets or exceeds `Llama 3.1 Instruct on Open Source LLM benchmarks`. Hermes 3 was powered by an 8-node `Lambda 1-Click Cluster` and features a uniquely unlocked, uncensored, and steerable model with advanced creative capabilities and strategic planning abilities. The model has been optimized for efficiency through data synthesis, supervised fine-tuning, reinforcement learning from human feedback, and quantization using Neural Magic's FP8 method. Hermes 3 can run on a single node or scale to a multi-node cluster for further fine-tuning, and is available for free through `Lambda's new Chat Completions API` and `Lambda Chat interface`.