AIAP Tool Guides·Google AI Studio·15:29

AI Studio Chat, Compare Mode and Temperature: Reading the Thinking Trace, Picking a Model, and Turning Creativity Up or Down

Karan Rana Brand/marketing manager, YouTuber, podcaster and course creator ("a non-tech guy"); presents the AI for Marketing, NotebookLM and Manus tool guides

▶LMS lesson

The short version

Gemini 2.5 Pro shows a collapsible "thinking" trace (about 22 seconds) before three viral reel hooks for an iPhone 16 India launch (0:02-0:05).
Compare mode runs one prompt in two models side by side; hover cards show pricing and use cases. 2.0 Flash-Lite answered first and short; 2.5 Pro was slower but better (0:05-0:12).
Temperature runs 0 to 2: 0 follows the prompt "to the T"; higher adds randomness - low for CEO decks, a little higher for Gen Z posters (0:12-0:15).

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01Compare mode: the same prompt, two modelsSplit the chat, choose a model per pane, run one prompt, judge speed versus quality.›

Split the chat, choose a model per pane, run one prompt, judge speed versus quality.

Thinking trace (0:02-0:05)

Compare (0:05-0:12)

↓ Full write-up of this concept

02Temperature: from literal to wildTemperature (labelled creativity) ranges 0 to 2;›

Temperature (labelled creativity) ranges 0 to 2; low values follow the prompt closely, high values add randomness and absurdity.

Temperature (0:12-0:15)

↓ Full write-up of this concept

The concepts in full

Compare mode: the same prompt, two models

how-to

"While 2.0 Flash-Lite was really good at coming up with answers quicker, the quality - 2.5 Pro won."

Hover cards at recording: 2.5 Pro for coding and reasoning, 2.5 Flash for agentic use, 2.0 Flash for native tool use, with per-token API pricing. On "give me a LinkedIn post caption on Oscars versus IIFA" (funny), Flash-Lite gave a tweet-length line; 2.5 Pro wrote a longer corporate-awards parody after ~22 seconds. Reading the thinking trace ("considering real hooks, developing potential hooks...") teaches how the model builds ideas.

Do it in this order

01Chat > two-arrow icon > Compare mode.
02Pick a model in each pane (hover for cards).
03Run one prompt; compare speed and quality.

Why it matters

Choose models by evidence on your own prompt.

This idea elsewhereparallelGenspark: choosing the model for the job

Temperature: from literal to wild

Low for the CEO proposal, a little higher for the Gen Z poster.

At 0 "the prompt will be followed word by word to the T without any randomness"; at 2 an exam-prep prompt might drift into fantasy. No default value is stated. Top-p, token counts, structured output and function calling are named but not taught; the transcript ends mid-explanation.

Why it matters

One setting that changes tone and reliability.

Tools referenced

Tool	Coverage	Moment	Context
Google AI Studio	demonstrated
Gemini 2.5 Pro	demonstrated
Gemini 2.0 Flash-Lite	demonstrated
Midjourney	mentioned
ChatGPT	mentioned
Claude	mentioned

Action items

Resources mentioned

Resources

docLesson page on the platform

Extraction notes

This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

Transcript corrections applied

The transcript says	The trainer actually means
real hooks	reel hooks
flashlight	Flash-Lite
IFA	IIFA
2.5 flash pro	2.5 Flash

True on recording day — verify before relying

Vendor UI walkthrough: menu names, layout, plan limits and pricing change monthly - check the live product before following click paths.
Model list, hover-card pricing and the thinking duration are June 2025; his claim that Midjourney has a "temperature" setting is unverified (0:07-0:13).