API reference
Choosing a model
A quick guide to choosing models by use case and budget. Refer to the Model Catalog (Models in the console) for capabilities and current prices.
Choose by use case
| Use case | Recommended models | Description |
|---|---|---|
| Everyday chat and content creation | deepseek-v4-flash, glm-5.2 | Fast responses at low cost |
| Software development | claude-sonnet-5, gpt-5.6-terra, deepseek-v4-pro | Code generation and reasoning |
| IDE code completion | gpt-5.6-terra | Specialized coding with strong value |
| Complex reasoning | claude-opus-4-8, gpt-5.5 | Mathematics, code and multi-step planning |
| Long-document processing | kimi-k3, gpt-5.6-sol | Context around 1M tokens |
| Chinese-language tasks and tool calling | glm-5.2, deepseek-v4-pro | Strong Chinese-language performance and value |
| Image and video understanding | claude-sonnet-5, gpt-5.6-sol | Multimodal visual input |
| Agents and automated workflows | claude-sonnet-5, gpt-5.6-sol | A mainstay for production agents |
| Image generation | See the Model Catalog for available models and the Media generation section for details | |
| Video generation | See the Model Catalog for available models and the Media generation section for details | |
Choose by budget
| Budget | Recommended models |
|---|---|
| Cost first | deepseek-v4-flash, glm-5.2, deepseek-v4-pro |
| Balanced | claude-sonnet-5, deepseek-v4-pro, glm-5.2 |
| Performance first | claude-fable-5, claude-opus-4-8, gpt-5.6-sol |
Selection approach
- Start small: Build the workflow with a lightweight model (such as
deepseek-v4-flash), then upgrade individual stages to a flagship where results are insufficient. - Compare results: Run the same real test cases on two or three candidates and compare quality and usage before selecting your main model.
- Choose context for the task: Longer context increases per-request consumption; not every task needs a very long context window.
- Mix models: Use different models for different stages, such as a lightweight model for classification and routing and a flagship for final generation.
Model prices, rate limits and route information are shown in real time in the Model Catalog. Spending details are available under Usage in the User Center.