ai-sage/GigaChat3.1-10B-A1.8B-bf16
Full fine-tune vs ai-sage/GigaChat3-10B-A1.8B-base. Fine-tune: 100% of the weights changed, across 27 of 27 layers.
Tensors changed: lm_head 1/1, Embeddings 2/2, Norms 85/85, MLP 2802/2802, Attention 108/108, other 2/2
7 derivatives checked
- ENOSYS/GigaChat3.1-10B-A1.8B-750-v1-GGUF: Quantization
- mradermacher/GigaChat3.1-10B-A1.8B-bf16-i1-GGUF: Quantization
- mradermacher/GigaChat3.1-10B-A1.8B-bf16-GGUF: Same weights, converted
- RockTalk/GigaChat3.1-10B-A1.8B-MLX-4bit: Quantization
- ai-forever/SMITH-Exp: Full fine-tune
- dkagramanyan/GigaChat3.1-10B-A1.8B-W4A16: Quantization
- blascotobasco/GigaChat3.1-Lightning-Uncensored: LoRA-style
3 other repos hold byte-for-byte identical weight files.