fastllm is a developer engineering workflows repository at ztxz16/fastllm; maintainers describe it as: 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+. Its recorded primary language is C++. License metadata lists Apache-2.0. GitHub metadata shows about 4,712 stars.
License
Apache-2.0
Stars
4,825
Features
- Recorded summary for fastllm: 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+。
- fastllm uses C++ as its recorded primary language, which helps with stack-fit review.
- fastllm fits engineering teams assessing code, CLI, SDK, runtime, or developer-tooling workflows.
- fastllm lists Apache-2.0 license metadata; review obligations before redistribution or hosted use.
- fastllm has about 4,712 GitHub stars in the local metadata snapshot.
- Repository identity: ztxz16/fastllm.
Use Cases
- Evaluate fastllm when the need is developer engineering workflows and the repo summary matches: 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...
- Compare the C++ implementation in fastllm before choosing a similar internal architecture.
- Use fastllm to study developer-tooling implementation details before building internal workflows.
- Complete a Apache-2.0 license review before packaging fastllm into a commercial or hosted workflow.
- Use fastllm's GitHub traction as one input when prioritizing open-source evaluation.
FAQ
Start from the repository summary (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+。), then verify maintenance status, integration boundaries, and whether its developer engineering workflows focus matches the intended workflow. Repository: https://github.com/ztxz16/fastllm. Stars: about 4,712. License: Apache-2.0. Language: C++.
fastllm is best treated as a repository-level component or reference implementation for developer engineering workflows. Good evaluation scenarios include: Evaluate fastllm when the need is developer engineering workflows and the repo summary matches: 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... Compare the C++ implementation in fastllm before choosing a similar internal architecture. Use fastllm to study developer-tooling implementation details before building internal workflows.
