Best for complex software engineering, toolchain collaboration, and converting long-context specifications into plans. A strong primary choice for high-quality development.

An advanced reasoning model for professional developers, with strong coding, long-context, and autonomous-agent capabilities.

Strong frontend output, multilingual adaptation, and code-task orchestration; suitable for page generation and engineering assistance.

Can build complex Agent Harness workflows; suitable for Agent Teams, complex Skills, and productivity automation.

Suitable for tool-heavy agents, customer workflows, and complex tasks requiring high execution quality.

Deep-thinking by default, supporting multi-step tool calls and reasoning; suitable for complex logic, math, and code agents.

A flagship general-purpose model with balanced multimodal ability, complex reasoning, structured generation, and long-horizon task handling.

Strong multilingual ability with improved reasoning and coding, suitable for enterprise content and general productivity scenarios.

Balances generation quality and response speed; suitable for high-frequency tasks such as content creation, search recommendation, and data analysis.

Input multiplier 0.25. Designed for efficient coding and Agent workflows, with the strongest cost advantage.

Input multiplier 0.3. Upgraded multilingual programming ability, suitable for low-cost coding and Agent calls.

Input multiplier 0.5. Stable and efficient for general Q&A, daily Agent tasks, and lightweight coding.