Model Selection
Introduction
Plugin Model Overview The plugin currently supports Smart Routing, Speed Model,Deep Model, Expert Model, Basic Model, and Third-Party Models for your selection.Smart Routing can intelligently select the most suitable model to complete tasks, improving efficiency and reducing costs. The Speed Model excels at core feature implementation, code refactoring,and daily development tasks, offering ultimate inference speed and ultra-fast response. The Deep Model specializes in architecture design and core feature implementation, with deep reasoning capabilities. The Expert Model is proficient in complex system design and handling high- difficulty problems, featuring expert-level reasoning and thinking abilities. The Basic Model is good at basic code generation and daily Q&A, with standard reasoning ability, while consuming fewer tokens and having lower costs. Please choose the appropriate model according to your task requirements to complete subsequent tasks. Please note that these models are only available to subscribers of the Personal Professional Edition.
Model Selection
- In Smart Guide, AI Toolbox, and Smart Chat, you can choose to use "Smart Routing", "Express Model", "Deep Model", or "Expert Model" based on the scenario and requirements of your task. You can also choose whether to enable "Thinking Mode".
| Dimension | Low | Medium | High | Higher | Highest |
|---|---|---|---|---|---|
| Analogy | "Intuitive reaction" like daily chatting, blurting out. | "Normal problem-solving" like doing homework, with steps but not tedious. | "Expert seminar" like scientific research, repeated deduction and self-correction. | "Top-tier think tank" like interdisciplinary strategic decision-making, exhausting all global variables. | "Ultimate deduction" like simulating parallel universes, exhausting all possibilities. |
| Speed | Extremely fast | Moderate | Relatively slow | Extremely slow | Extremely slow (approaching system latency limits) |
| Applicable Scenarios | Casual chatting, translation, simple Q&A, creative brainstorming. | Writing emails, summarizing articles, routine programming, general logic puzzles. | Complex mathematics, advanced code architecture, legal analysis, scientific reasoning. | Complex system design, interdisciplinary research, strategic business decision-making. | Frontier scientific exploration, global crisis response, ultimate philosophical speculation. |
| Advantages | Rapid response, low cost, fluent and natural. | Balances speed and accuracy, high cost-performance ratio. | Extremely high accuracy, capable of handling multi-step complex logic, reduces hallucinations. | Capable of handling ultra-large-scale complex systems, possesses global optimal solution capability. | Capable of exploring unknown territories, possesses potential for creative breakthroughs. |
| Disadvantages | Prone to errors, difficult to handle complex logical traps. | May lack rigor on extremely complex issues. | Time-consuming, high computational resource consumption. | Extremely high resource consumption, significantly prolonged response time. | Nearly infinite resource consumption, may exceed actual application scenario requirements. |
| Description | Minimal reasoning, fastest response | Moderate reasoning depth | In-depth reasoning, suitable for complex tasks | Deep analysis, applicable to highly difficult problems | Maximum reasoning depth, suitable for the most complex challenges |
- The model will provide answers based on the mode you selected.
Smart Routing
Intelligently selects the most suitable model based on the task. For most daily development work, you can choose "Smart Routing" to improve efficiency and reduce costs.
Express Model
The Express Model excels at core feature implementation, code refactoring, and daily development tasks, offering ultimate inference speed and lightning-fast responses.
Deep Model
The Deep Model specializes in architecture design and core feature implementation, providing deep reasoning capabilities.
Expert Model
The Expert Model is designed for complex system design and handling highly difficult problems, possessing expert-level reasoning and thinking abilities.
Please note: When switching to the Expert model, due to its high computational demands, token usage will be consumed at a 5x multiplier. Please choose carefully.
Basic Model
Excels at basic code generation and daily Q&A, with standard reasoning capabilities, while offering low token consumption and low cost.
Third-Party Models
Models from third-party channels are integrated. Please select the appropriate third-party model based on your requirements.
- GLM-5.2: Designed for long-horizon tasks, capable of autonomous planning and execution with continuous evolution, delivering engineering-grade results.
- GLM-5.3: Code capabilities benchmark against top international standards, featuring both security review and vulnerability reasoning capabilities.
- DeepSeek-V4-Pro-0813: Primarily targeted at high-complexity, multi-step autonomous task scenarios.
- DeepSeek-V4-Flash-0731: Specifically optimized for Agent scenarios (especially code development and tool invocation).
- DeepSeek-V4.1-Flash: Comprehensive leadership in Agent capabilities, world knowledge, and reasoning performance.
- Step-5-Preview: StepFun's new-generation flagship multimodal reasoning model.
