SenseCore

What it does

SenseCore is SenseTime's AI infrastructure platform, combining compute services, a model layer, and development tools for training and deploying AI workloads. Its cloud offerings include GPU and CPU cloud servers, bare-metal GPU servers, Kubernetes-managed training through ECP, and serverless container instances for online services. Its model-as-a-service layer provides access to the SenseNova large model family, which covers language, multimodal, speech, and embedding models, along with plugins such as web search and a RAG knowledge base.

Teams typically use it for distributed model training and inference. The ACP compute platform supports PyTorch and MPI, large-scale parallel training, fault detection and recovery, and monitoring, while ECP adds resource management and checkpoint saving for training and deployment. Delivery is available as a product licence or a cloud service, with expert services covering consulting, model training, and inference deployment.

SenseCore suits enterprise and research teams that need large-scale GPU capacity in China and are comfortable working through SenseTime's sales and support channels. Access is through the console and cloud services, and the help centre covers account creation, user authorization, and resource management. Compute is hosted in SenseTime facilities rather than on local devices, and much of the documentation is in Chinese.

What it can do

• Large-scale distributed AI model training with fault recovery and checkpoint saving
• GPU and CPU cloud server lifecycle management
• Kubernetes-based training and resource management (ECP)
• Serverless container instances for online services with health checks and elastic scaling
• Bare-metal GPU servers with physical isolation
• Model-as-a-service access to the SenseNova large model family

Works with

• PyTorch
• MPI

Example workflows

Distributed model training

Input: a PyTorch training job and a GPU cluster allocation. Steps: submit the job through the Kubernetes-managed platform, enable checkpoint saving and monitoring, and let the scheduler recover from node failures. Output: a trained model with checkpoints and monitoring data.

Deploying an online service

Input: a containerized inference service. Steps: deploy it to serverless container instances with health checks and elastic scaling. Output: a running online service with monitoring.

Before you choose

• Hosted compute runs in SenseTime's China-based facilities, which affects data residency and compliance planning.
• Much of the material is in Chinese; confirm language support and documentation for your team.
• Check current pricing with SenseTime, since costs depend on the product, usage, and delivery mode.

Listed under

• Coding
• Sandboxes and infrastructure
• AI cloud
• AI infrastructure
• Computing platform
• Deep learning
• Deep learning platform
• Enterprise AI

Get started

A quick purchase flow and user setup are described in the help centre.

1. Create user accounts and authorize access using the help centre's user management guides.
2. Review resource management and choose the appropriate compute product (ECS, BMS, ACP, or CCI).
3. Complete the quick purchase flow or contact sales to set up billing and delivery mode.

Billing is usage- or deployment-based and set through SenseTime's sales channels.

We couldn’t find published prices for SenseCore. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• SenseCore 商汤大装置-AI基础设施新范式 ↗www.sensecore.cn
• 了解我们 ↗www.sensecore.cn
• 大装置帮助中心 ↗www.sensecore.cn

Company website

sensecore.cn: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.