home/coding/gemini3-family-prompting

Gemini3 Family Prompting

GPTClaudeDeepSeek··▲ 1,308 copies·updated 2026-07-14

gemini3-family-prompting.prompt

# Gemini 3 Family Prompting Guide

Covers Gemini 2.5 Pro/Flash and Gemini 3/3.1 Pro. Based on official Google AI documentation and production patterns.

Sources:
- [Gemini 3 Prompting Guide — Vertex AI](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/start/gemini-3-prompting-guide)
- [Gemini API Prompting Strategies](https://ai.google.dev/gemini-api/docs/prompting-strategies)
- [Gemini 3 Prompt Practices — Philipp Schmid](https://www.philschmid.de/gemini-3-prompt-practices)
- [Gemini Thinking Mode](https://ai.google.dev/gemini-api/docs/thinking)
- [Function Calling — Gemini API](https://ai.google.dev/gemini-api/docs/function-calling)
- [Structured Output — Gemini API](https://ai.google.dev/gemini-api/docs/structured-output)
- [Gemini API Cookbook](https://github.com/google-gemini/cookbook)

## Model Comparison

| Aspect | Gemini 2.5 Flash | Gemini 2.5 Pro | Gemini 3 Pro | Gemini 3.1 Pro |
|--------|-----------------|----------------|-------------|----------------|
| Thinking control | `thinking_budget` (tokens) | `thinking_budget` (tokens) | `thinking_level` (LOW/HIGH) | `thinking_level` (LOW/HIGH) |
| Default thinking | Dynamic (-1) | Dynamic (-1) | HIGH | HIGH |
| Context window | 1M tokens | 1M tokens | 1M tokens | 1M tokens |
| Key strength | Speed + image gen | Complex reasoning | Direct instruction following | 2x abstract reasoning vs 3.0 |
| Verbosity | Normal | Normal | Less verbose by default | Less verbose by default |
| Temperature | 1.0 (recommended) | 1.0 (recommended) | **1.0 (critical)** | **1.0 (critical)** |

## Critical: Temperature for Gemini 3

**Keep temperature at 1.0.** Lowering temperature below 1.0 may lead to unexpected behavior, looping, or degraded performance, particularly with complex mathematical or reasoning tasks.

This is a significant difference from other model families where lower temperature = more deterministic. Gemini 3's reasoning is optimized specifically for temperature 1.0.

## Thinking Mode Configuration

### Gemini 2.5 (Token-Based Budget)

when to use it

Community prompt sourced from the open-source GitHub repo CodeAlive-AI/ai-driven-development (MIT). A "Gemini3 Family Prompting" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.