NousResearch/Hermes-4-405B
Full fine-tune vs meta-llama/Llama-3.1-405B. Fine-tune: 100% of the weights changed, across 126 of 126 layers.
Tensors changed: Embeddings 1/1, Attention 504/504, MLP 378/378, Norms 252/253, lm_head 1/1
9 derivatives checked
- bartowski/NousResearch_Hermes-4-405B-GGUF: Quantization
- DevQuasar/NousResearch.Hermes-4-405B-GGUF: Same weights, converted
- unsloth/Hermes-4-405B-GGUF: Same weights, converted
- lmstudio-community/Hermes-4-405B-MLX-4bit: Quantization
- lmstudio-community/Hermes-4-405B-MLX-8bit: Quantization
- lmstudio-community/Hermes-4-405B-MLX-6bit: Quantization
- lmstudio-community/Hermes-4-405B-MLX-5bit: Quantization
- lmstudio-community/Hermes-4-405B-GGUF: Quantization
- mradermacher/Hermes-4-405B-i1-GGUF: Not comparable yet
4 other repos hold byte-for-byte identical weight files.