DeepSeek-V3.2-Exp#

  • Context Length: 163840

  • Model Name: DeepSeek-V3.2-Exp

  • Languages: en, zh

  • Abilities: chat, reasoning, hybrid, tools

  • Description: We are excited to announce the official release of DeepSeek-V3.2-Exp, an experimental version of our model. As an intermediate step toward our next-generation architecture, V3.2-Exp builds upon V3.1-Terminus by introducing DeepSeek Sparse Attention—a sparse attention mechanism designed to explore and validate optimizations for training and inference efficiency in long-context scenarios.

Specifications#

Model Spec 1 (pytorch, 671 Billion)#

  • Model Format: pytorch

  • Model Size (in billions): 671

  • Quantizations: none

  • Engines: vLLM, Transformers

  • Model ID: deepseek-ai/DeepSeek-V3.2-Exp

  • Model Hubs: Hugging Face, ModelScope

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name DeepSeek-V3.2-Exp --size-in-billions 671 --model-format pytorch --quantization ${quantization}

Model Spec 2 (awq, 671 Billion)#

  • Model Format: awq

  • Model Size (in billions): 671

  • Quantizations: AWQ, AWQ-Lite

  • Engines: Transformers

  • Model ID: QuantTrio/DeepSeek-V3.2-Exp-{quantization}

  • Model Hubs: Hugging Face, ModelScope

Execute the following command to launch the model, remember to replace ${quantization} with your chosen quantization method from the options listed above:

xinference launch --model-engine ${engine} --model-name DeepSeek-V3.2-Exp --size-in-billions 671 --model-format awq --quantization ${quantization}