Gemini 3.1 Flash-Lite (Preview)

Overview

Gemini 3.1 Flash-Lite is Google's most cost-effective multimodal model. It is built for the fastest performance on lightweight, high-frequency workloads, making it a strong fit for large-scale agent tasks, simple data extraction, translation, transcription, and ultra-low-latency applications where budget and speed matter most.

Best Use Cases

✅ Recommended For

• High-Volume Agents: Lightweight routing, classification, and orchestration at scale
• Translation: Fast and inexpensive multilingual processing
• Transcription: Audio-to-text workflows without a separate speech pipeline
• Structured Extraction: Entity extraction, classification, and JSON outputs from text or documents
• Document Summaries: Quick summaries over PDFs and mixed multimodal inputs

⚠️ Consider Alternatives For

• Complex Reasoning: Gemini 3.1 Pro is better for deeper multi-step reasoning
• Advanced Coding: Gemini 3 Flash or 3.1 Pro are better choices for heavier coding tasks

Technical Highlights

Feature	Details
Context Window	1,048,576 input tokens
Max Output	65,536 output tokens
Pricing	$0.25 / $1.50 per MTok (text, image, video input / output), audio input at $0.50 per MTok
Thinking	✅ Supported with thinkingLevel (minimal, low, medium, high), dynamic high by default
Multimodal Input	✅ Text, image, video, audio, and PDF
Tools	✅ Function calling, structured output, Google Search grounding, URL context, file search, and code execution
Cache & Batch	✅ Context caching and Batch API supported
Streaming	✅ Full support
Not Supported	No Live API, computer use, Google Maps grounding, image generation, or audio generation