Model ID
deepseek-ai/DeepSeek-V3.2-Exp
Capabilities
DeepSeek V3.2 Exp
DeepSeek V3.2 Exp is an experimental Mixture-of-Experts (MoE) large language model developed by DeepSeek AI as an intermediate research step toward its next-generation architecture. It builds upon the V3.1-Terminus design while introducing DeepSeek Sparse Attention (DSA). The model features a 685B parameter architecture with ~37B activated parameters, built on a 61-layer transformer with 128 attention heads and a 7,168 hidden size, utilizing 256 routed experts with 8 experts activated per token. It supports a ~160K token context window using YaRN-based rotary scaling.
About this model
DeepSeek V3.2 Exp is an experimental Mixture-of-Experts (MoE) large language model developed by DeepSeek AI as an intermediate research step toward its next-generation architecture. It builds upon the V3.1-Terminus design while introducing DeepSeek Sparse Attention (DSA). The model features a 685B parameter architecture with ~37B activated parameters, built on a 61-layer transformer with 128 attention heads and a 7,168 hidden size, utilizing 256 routed experts with 8 experts activated per token. It supports a ~160K token context window using YaRN-based rotary scaling.
Model specifications and capabilities are published by the model author and reproduced here from HF Model Card (deepseek-ai/DeepSeek-V3.2-Exp). Released under MIT. Further reading: Paper · Blog
Deploy Private Inference Endpoint
Choose your hardware and inference engine to get deployment commands and performance benchmarks tailored to your infrastructure
Hardware
MI325X
2,048 GB
B200
1,440 GB
Engine
vLLM
SGLang
Deployment code
Access Restricted
Enter your email to get access to this content
Performance under load
Tokens/s (total)
Benchmarks measured by Vultr on MI325X · vLLM. Throughput and latency vary with concurrency, input length and engine configuration, so treat these as a comparison baseline, not a service guarantee.
AMD Instinct MI325X
Unlock Next-Generation AI Power on Vultr Cloud GPU