fastllm OpenAPI & REST API Specification
Fastllm是后端无依赖的高性能大模型推理库。同时支持张量并行推理稠密模型和混合模式推理MOE模型，任意10G以上显卡即可推理满血DeepSeek。双路9004/9005服务器+单显卡部署DeepSeek满血满精度原版模型，单并发20tps；INT4量化模型单并发30tps，多并发可达60+。
5,112 stars
C++
1 file · ~126,242 tokens
No Explicit API Specification Located
This repository does not currently contain a static OpenAPI, Swagger, or GraphQL schema file. You can still access its complete technical documentation or agent skill rules below: