Reported impact
Business outcome
Changan significantly increased GPU efficiency, accelerated intelligent-vehicle development, and established a long-term AI base for both enterprise and in-vehicle innovation.
- Productivity Improvement
- Scalability
Baidu AI CloudAutomotiveAI and Machine LearningCloud NativeData PlatformProductivity ImprovementScalability2025 Baidu AI CloudBD#2932e1AutomotiveChinaChangan needed a large-scale intelligent-computing foundation to support growing data volumes and accelerate work on autonomous driving, connected vehicles, and large-model applications as it transformed into an…142 exa-operations per second compute capacityMay 14, 20256Baige AI Heterogeneous Computing Platform · ERNIE · Xinghuan Platform0 2025-05-14 6
Changan Automobile (长安汽车)
Changan needed a large-scale intelligent-computing foundation to support growing data volumes and accelerate work on autonomous driving, connected vehicles, and large-model applications as it transformed into an intelligent low-carbon mobility company.
142 exa-operations per second compute capacity
30,000+ AI model-training runs completed
GPU utilization improved by 40%+
nearly 100 million labeled frames accumulated
Solution
Baidu AI Cloud and Changan built an intelligent-computing center around the Baige AI heterogeneous computing platform and Changan's own Xinghuan platform. The compute center delivers 14.2 exa-operations per second and gives Changan a unified environment that spans data collection, processing, labeling, training, evaluation, and model deployment for intelligent-driving and connected-vehicle programs. Baige acts as the infrastructure base, while Xinghuan provides unified scheduling and management across clusters, GPU resources, and storage. The platform has already accumulated nearly one hundred million labeled frames and more than 30,000 AI model-training runs, which means Changan can treat the environment as a reusable industrial AI factory rather than a point solution. Baidu also extends the same foundation into ERNIE-powered in-vehicle experiences, enterprise knowledge-management experiments, and compute-optimization services for Changan's self-developed industry models, making the intelligent-computing center a shared base for both vehicle-facing and enterprise-facing AI.
Data flow
Vehicle and R&D data are collected, processed, labeled, trained, evaluated, and deployed through Xinghuan on top of Baige's compute center, while optimized models are used for intelligent-driving, connected-vehicle, knowledge-management, and in-vehicle dialogue scenarios.
Solution architecture
- Compute
- Baige AI Heterogeneous Computing Platform Provides the intelligent-computing infrastructure base for Changan's model training, connected-vehicle analytics, and autonomous-driving R&D.
- Serving
- ERNIE Supports in-vehicle dialogue and service capabilities as well as enterprise knowledge-management experiments on top of the intelligent-computing center.
- Orchestration
- Xinghuan Platform Handles end-to-end data and model workflows from collection and labeling through training, evaluation, deployment, and unified cross-cluster resource scheduling.
Architecture clues
- Baidu provides additional large-model compute optimization for Changan's self-developed industry models.
- The intelligent-computing center supports both automotive R&D and enterprise AI applications.
- The platform unifies data collection, labeling, training, evaluation, deployment, and resource scheduling rather than treating those as separate toolchains.
Evidence from the source
Baidu and Changan are also using ERNIE in vehicle models and enterprise knowledge-management scenarios. (original: 在车型中搭载百度文心一言大模型...在企业知识管理方面，探索应用文心大模型。)
Changan's intelligent-computing center is based on Baidu AI Cloud's Baige AI heterogeneous computing platform. (original: 长安汽车和百度共建的长安汽车智算中心基于百度智能云的百舸·AI异构计算平台。)
System launch improved GPU resource utilization by more than 40%. (original: 系统上线以来，GPU（图形处理器）资源利用率提升40%以上。)
The center supports the full workflow from data collection, processing, labeling, training, and evaluation to model deployment through the Xinghuan platform. (original: 打造了支持从数据采集、处理、标注、训练、评测到模型部署全流程的“星环平台”。)
The platform has already accumulated nearly one hundred million high-quality labeled frames and completed more than 30,000 intelligent AI algorithm training jobs. (original: 已累积近亿帧的高质量标注数据，累计完成超3万次的智能AI算法模型训练。)
Transparency Evidence & methodology
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
How scores are calculated
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Customer reference value100/100
- Problem clarity20/20Measures whether the customer's operational problem is specific enough to compare.
- Implementation depth25/25Rewards named components, data flow, integrations, and architecture detail.
- Outcome specificity20/20Measures whether the reported benefit is concrete rather than a generic claim.
- Quantified impact25/252 metrics describe measurable impact.
- Deployment scale10/10Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence89%
- Source authority20/25The main evidence is an official vendor-published customer story.
- Claim traceability21/2510 of 12 key fields link to a named source.
- Direct evidence20/205 direct source quotes retained.
- Date provenance10/10The publication date comes from the source page or publisher.
- Source transparency8/101 source reference document publisher, URL, authority, retrieval, and supported fields.
- Implementation specificity10/10Rewards source-backed components, data flow, quantified impact, and a specific outcome.
Traceability
Sources & provenance
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
- https://cloud.baidu.com/customer/case/changancar.htmlBaidu AI CloudFirst-party
Changan Automobile (长安汽车) customer story
Supports- Customer
- Business problem
- Solution
- Products
- Solution components
- Data flow
- Integrations
- technical area
- use case category
- Business outcome
- outcome category
- Metrics
- Architecture
- evidence quotes
- Publication date
Date evidence
- Published on source page May 14, 2025 · Baidu AI Cloud · First-partyOpen date source
Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.
Related
Similar use cases
Baidu AI Cloud Sample Jan 8, 2026 Mercedes-Benz Group AG Mercedes-Benz needed to unify fragmented customer-contact operations spread across assistant, telematics, and roadside-assistance centers so it could improve oversight, expand service capacity, and deliver a more… End-to-end online Q&A accuracy exceeded 85%. ASR · Baidu AI Cloud Intelligent Customer Service · NLG Baidu AI Cloud Sample Jun 27, 2025 SAIC Volkswagen Automotive Co., Ltd. SAIC Volkswagen needed to support multiple service and marketing departments, headquarters, regions, and dealers from one customer-direct-contact center while expanding dual-brand operations, multichannel access, and… ASR recognition accuracy reached 90%. ASR · Baidu AI Cloud Intelligent Customer Service · CDCC Customer Direct Contact Center Baidu AI Cloud Sample Jun 20, 2025 Li Auto Li Auto needed stable, high-efficiency cloud services to support fast-growing in-vehicle digital content, intelligent-cockpit interaction, and always-on vehicle services without compromising safety, user experience, or… BCC · Qianfan AppBuilder · Qianfan ModelBuilder