Guide · Article 03
Choosing a reasoning level
When a higher reasoning level is needed, and when a low level is enough to stay fast and economical.
F68 can display these levels:
Low · Medium · High · XHigh · Max
Not every model supports all levels. F68 uses the level that matches what each model can actually do.
Higher reasoning levels usually take more time and cost more.
It is not a “make the answer better” button for every situation.
LOW
Light reasoning.
Best for:
- chat;
- email;
- summarization;
- rewriting;
- small coding tasks;
- everyday work.
MEDIUM
Balances:
quality · speed · cost
Best for:
- research;
- analysis;
- writing;
- coding;
- planning;
- Agents;
- knowledge base.
If you don’t know which level to choose, start with Medium.
HIGH
For work that needs deep reasoning.
Best for:
- in-depth analysis;
- complex research;
- complex coding tasks;
- debugging;
- important planning;
- critique;
- finance;
- multi-step workflows.
XHIGH
For especially demanding tasks.
Best for:
- complex coding;
- long-horizon Agents;
- hard research;
- problems with many conditions;
- cases where High isn’t enough.
Not recommended as a default.
MAX
The highest reasoning level, when the model supports it.
Best for:
- exceptionally hard problems;
- multi-round research;
- complex coding or system architecture;
- long-running agent workflows with many steps and constraints;
- cases where XHigh is still not enough.
Max is usually the slowest and most expensive. Not recommended as a default.
Not every model supports Max. F68 AI only shows or uses this level when the model actually supports it.
Quick way to choose
- Simple work
- Economical model + Low
- Everyday professional work
- Balanced model + Medium
For example:
Terra · Sonnet · Flash · Grok
- Hard work
- Strong model + High
- Exceptionally hard or important work
- Flagship + XHigh + critique by another model
- Extremely hard, long or multi-round work
- Flagship + Max + critique by another model
Some real-world situations
- Writing an email
- Luna / Haiku + Low
No flagship needed.
- Writing an analysis report
- Terra / Sonnet + Medium
For more demanding reports, increase the reasoning level to High.
- Important financial research
- Opus / Fable / Sol + High
Cross-check with another model.
- Project planning
- Terra / Sol + Medium or High
- Everyday coding
- Sonnet / Terra / GLM + Medium
- Debugging a complex system
- Opus / Fable / Sol + High
- Reading a very large volume of documents
- Gemini / Kimi + Medium or High
- Processing thousands of identical tasks
- Luna / Haiku / Flash / DeepSeek / GLM / Qwen + Low
Don’t use a flagship for every single task.