Start with Flash
Use the balanced model for most workplace requests and review the first response.
Gemini models · speed and reasoning
Choose a Gemini model
Gemini offers three main model choices for different balances of speed and reasoning. Start with Flash for normal work and choose Pro when the task is genuinely difficult.
Availability and usage limits depend on your account, plan and current capacity.
The three Gemini model choices
💨Flash-LiteQuick summaries, brainstorming and low-complexity repeated work.
⚡FlashThe balanced default for most professional requests.
🧠ProDifficult analysis, complex files and demanding multimodal work.
Match the model to the work
Work
Recommended start
Move up when
Routine summary or rewrite
Flash-Lite or Flash
The material is unusually long or ambiguous.
Spreadsheet, report or file analysis
Flash
Several files or conflicting assumptions must be connected.
Difficult logic or technical reasoning
Pro
The problem genuinely benefits from more reasoning time.
Mixed text, images and video
Pro
Improve the source material before assuming a model change will solve missing context.
Escalate only for a clear reason
Diagnose the weakness
Decide whether the problem is reasoning, missing context, unclear instructions or poor source material.
Move to Pro when complexity is real
Use stronger reasoning when Gemini must connect several constraints, files or media types.
Deep Think is a Pro setting
It is not a separate model
Google describes Deep Think as an experimental thinking level chosen after selecting Pro.
It may take several minutes
Use it only for unusually difficult problems where extra reasoning time is worth the wait.
It still needs review
Deep Think can still make mistakes and has separate eligibility and usage limits.
Sometimes you need a different feature
Deep Research
Use it for a structured investigation across current sources.
Notebook
Use it when project sources and instructions should persist.
Gem
Use it when the same role, method or output style repeats.
Change the model in chat
At the bottom of the Gemini chat box, select the current model name and choose Flash-Lite, Flash or Pro. On eligible accounts, selecting Pro and opening the model menu again reveals the optional Deep Think thinking level.
If an answer is weak, improve the request or provide better context before assuming a more expensive reasoning setting is necessary.
Practical rule
Flash-Lite, Flash and Pro are model choices. Deep Think is an optional experimental thinking level within Pro, not a fourth model.