NousResearch/Hermes-4-70B
Full fine-tune vs meta-llama/Llama-3.1-70B. Fine-tune: 100% of the weights changed, across 80 of 80 layers.
Tensors changed: Embeddings 1/1, Norms 160/161, MLP 240/240, Attention 320/320, lm_head 1/1
15 derivatives checked
- lmstudio-community/Hermes-4-70B-MLX-4bit: Quantization
- lmstudio-community/Hermes-4-70B-MLX-8bit: Quantization
- lmstudio-community/Hermes-4-70B-MLX-6bit: Quantization
- lmstudio-community/Hermes-4-70B-MLX-5bit: Quantization
- lmstudio-community/Hermes-4-70B-GGUF: Quantization
- unsloth/Hermes-4-70B-GGUF: Same weights, converted
- bartowski/NousResearch_Hermes-4-70B-GGUF: Quantization
- mradermacher/Hermes-4-70B-i1-GGUF: Quantization
- cyankiwi/Hermes-4-70B-AWQ-4bit: Quantization
- DevQuasar/NousResearch.Hermes-4-70B-GGUF: Quantization
- mradermacher/Hermes-4-70B-GGUF: Quantization
- genevaquant/Hermes-4-70B-GGUF-u: Same weights, converted
- mlx-community/Hermes-4-70B-8bit: Quantization
- mlx-community/Hermes-4-70B-4bit: Quantization
- AlSamCur123/Hermes-Llama3.1-70bContinued: LoRA-style
8 other repos hold byte-for-byte identical weight files.