Guides
Practical, sourced guidance for evaluating and using Chinese AI models, APIs and agents. Pillars are multi-chapter arguments; the guides below them are self-contained.
Buying Chinese AI: Access, Contracts and Vendor Risk
Chinese models are easy to try and hard to buy. A procurement pillar on payment rails, contracting entities, licence terms, data residency and vendor risk.
Start here · 8 chapters- How Foreign Teams Pay for Chinese AI APIs — chapter 1
- Who You Actually Sign With at Chinese AI Labs — chapter 2
- What Chinese AI Terms Actually Restrict — chapter 3
- Data Residency and Cross-Border Transfer for Chinese AI — chapter 4
- How to Spot a Chinese AI API Reseller — chapter 5
- Enterprise Onboarding at Chinese AI Vendors — chapter 6
- Self-Hosting vs API: A Cost Model That Survives Utilisation — chapter 7
- A Vendor Risk Register for Chinese AI Suppliers — chapter 8
Choosing a Chinese AI Model by Workload
A selection method that starts from the task, not the leaderboard: five filters in order, the six workloads that differ, and when a cheap model is the wrong answer.
Start here · 8 chapters- Choosing a Chinese Model by Price: Cache and Tiers — chapter 1
- Long-Context Model Selection: What 1M Tokens Really Mean — chapter 2
- Choosing a Multimodal Chinese Model: Vision and Video — chapter 3
- Choosing a Chinese Coding Model: Evidence, Not Claims — chapter 4
- Choosing a Chinese AI Agent: Models, Tools, MCP — chapter 5
- Choosing a Model for Document Extraction — chapter 6
- Choosing a Model for High-Volume Classification — chapter 7
- Build a 50-Item Eval Before You Choose — chapter 8
All guides
- Choosing a Chinese AI API Platform: Five Checks
Five checks before committing to a Chinese AI API platform: compatibility, pricing structure, region access, rate limits and contract terms.
- Enterprise Deployment of Chinese AI: Regions and SLA
Enterprise deployment of Chinese AI models: region availability, data residency, SLA terms and the procurement questions that decide feasibility.
- Self-Hosting Chinese Open Models: Licences and Hardware
Self-hosting Chinese open-weight models: licence restrictions, GPU memory requirements, quantization choices and the real cost per million tokens.
- Open-Weight vs API: Trade-offs for Chinese Models
Open-weight versus API access for Chinese models: licence rights, hardware cost, operational burden and the volumes at which self-hosting pays.
- How to Read Vendor-Reported AI Benchmarks
How to read vendor-reported benchmark numbers: which settings inflate scores, what is missing from the table, and when a claim cannot be compared at all.