Gemini 3.1 Flash-Lite

// Overview

Gemini 3.1 Flash-Lite is Google's fastest and lowest-cost member of the Gemini 3 family, built for high-volume, latency-sensitive jobs such as translation, classification, and large-scale summarisation. It is a natively multimodal reasoning model based on Gemini 3 Pro, accepting text, image, audio, and video input with a 1M-token context window and up to 64K output tokens.

Google launched Gemini 3.1 Flash-Lite in preview on March 3, 2026, and made the GA model (id gemini-3.1-flash-lite) generally available on May 7, 2026; the earlier gemini-3.1-flash-lite-preview endpoint was shut down on May 25, 2026. The model is served through the Gemini API, Google AI Studio, Vertex AI, the Gemini app, and Search AI Overviews.

Despite the low price, Google positions Flash-Lite as a capable model: at launch it reported the top score across six of the benchmarks Google used to compare it with GPT-5 mini and Claude 4.5 Haiku. It also runs noticeably faster than the previous generation — Google cites roughly 45% higher overall generation speed and a time-to-first-token about 2.5x shorter than Gemini 2.5 Flash.

// Benchmarks

1. GPQA Diamond86.9%
2. MMMLU88.9%
3. MMMU-Pro76.8%
4. LiveCodeBench72%
5. Video-MMMU84.8%
6. Humanity's Last Exam16%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

Standard tier, text/image/video; audio input is $0.50/1M. Context-cache storage $1.00/1M tokens per hour.

Pricing source ↗

// Strengths

• Very low cost at $0.25 input / $1.50 output per million tokens, aimed at intelligence per dollar
• Fast: ~363 tokens/second output, ~45% faster generation and ~2.5x shorter time-to-first-token vs Gemini 2.5 Flash
• 1M-token context window for long documents and large multimodal prompts
• Natively multimodal — accepts text, image, audio, and video input
• Strong reasoning for its tier: 86.9% on GPQA Diamond and 88.9% on MMMLU

// Best for

• Reach for it for high-volume, cost-sensitive workloads like translation, classification, and tagging.
• Reach for it when you need low latency and fast first-token response in user-facing apps.
• Reach for it to process long or multimodal inputs (audio/video/images) cheaply within a 1M-token window.

// How to access

// Gemini Flash-Lite — every version

The full lineage of the Gemini Flash-Lite line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.