Reasoning

Reasoning is selected alongside the model and controls how much internal analysis Cosine asks that model to do while working on your request. It affects how carefully the agent compares options, chooses tools, handles ambiguity, and checks its work before responding or editing files.

What reasoning changes

Higher reasoning means:

• More planning before acting
• Better performance on ambiguous or multi-step tasks
• Slower responses
• More token usage

Lower reasoning means:

• Faster responses
• Less internal analysis
• Better fit for routine or well-scoped work

Reasoning levels

Level	Best for
none	Straightforward lookups, quick summaries, very simple edits
low	Small bug fixes, targeted searches, routine code changes
medium	Most day-to-day coding tasks (default)
high	Debugging tricky issues, design work, careful refactors
xhigh	Deep analysis on supported GPT/Codex models
adaptive	Supported Claude 4.6 models — lets the model choose how much thinking it needs

Practical advice

• Start with medium unless you have a reason not to
• Drop to low or none for repetitive, well-defined tasks
• Raise to high when the task is risky, ambiguous, or spread across several files
• Use adaptive on supported Claude 4.6 models when task complexity is likely to vary
• Use xhigh sparingly — best when the extra delay is worth the extra care

Setting reasoning

You can set reasoning per session or as a default:

cos start --reasoning high

Or in your config:

[inference]
reasoning_level = "medium"

Reasoning and cost

Reasoning tokens count toward total token consumption. Higher reasoning settings usually mean the model spends more tokens thinking before it answers, so the total cost of running that model goes up. See Model pricing for pricing details.

Related pages

• CLI Reasoning — Full reasoning reference with examples
• Model pricing — How reasoning affects cost