[Utilities] Cheat Sheet Internal Style Params of GPT

[umar-anzar]

Cheat Sheet Internal Style Params of GPT

Parameter	Typical Range / Example	What It Controls
Yap	1024 → 8192+	Maximum response length / token allowance.
oververbosity	0 → 10	How wordy or detailed responses are.
Personality	v1, v2, v3	Tone, empathy, humor, phrasing style.
Juice	0 → 10	Energy, assertiveness, confidence of answers.
Conciseness	0 → 10	Opposite of oververbosity; how tight and brief responses are.
Creativity / Randomness	0 → 10	Willingness to take imaginative or non-literal paths.
Safety / Content filter	low / medium / high	Strictness of content moderation.
Context window	1024 → 128k tokens	How much conversation history the model “remembers” in one session.
Memory mode	off / temporary / persistent	Whether the model retains info between sessions.
Tone adjustments	formal / casual / friendly	Minor tweaks to phrasing style beyond Personality.

💡 Quick mental takeaway:

• Length → Yap
• Wordiness → oververbosity / conciseness
• Style → Personality / Tone
• Energy → Juice
• Imagination → Creativity
• Safety → Content filter
• Memory → Context window / Memory mode