deepseek-ai/DeepSeek-V3.1-Terminus
Full fine-tune vs deepseek-ai/DeepSeek-V3.1-Base. Fine-tune: 100% of the weights changed, across 62 of 62 layers.
Tensors changed: Embeddings 2/2, Attention 578/620, Norms 252/252, MLP 2046/2123, lm_head 1/1, other 2/2
5 derivatives checked
- unsloth/DeepSeek-V3.1-Terminus-GGUF: Quantization
- bartowski/deepseek-ai_DeepSeek-V3.1-Terminus-GGUF: Quantization
- DevQuasar/deepseek-ai.DeepSeek-V3.1-Terminus-GGUF: Quantization
- tencent/DeepSeek-V3.1-Terminus-W4AFP8: Different architecture
- ling1000T/DeepSeek-V3.1-Terminus-gguf: Quantization