Gemini 3.1 Flash-Lite: Built for intelligence at scale

· Source: Google DeepMind News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, quick

Summary

Google introduced Gemini 3.1 Flashlight on March 3, 2026, as the fastest and most cost-efficient model in its Gemini 3 series, designed for high-volume developer and enterprise workloads. Priced at $0.25 per 1 million input tokens and $1.50 per 1 million output tokens, it offers enhanced performance and low latency, outperforming 2.5 Flash with 2.5 times faster time-to-first answer token and a 45% increase in output speed. The model achieved an ELO score of 1432 on Arena.AI and scored 86.9% on GPQA Diamond and 76.8% on MMUU Pro, surpassing prior-generation Gemini models. It also includes "thinking levels" for developers to control reasoning depth, making it suitable for tasks like high-volume translation, content moderation, UI generation, and simulations.

Key takeaway

For AI Engineers and CTOs managing high-volume, cost-sensitive AI deployments, Gemini 3.1 Flashlight presents a compelling option. Its low latency and competitive pricing of $0.25/M input tokens and $1.50/M output tokens, combined with strong benchmark performance, suggest it can reduce operational costs while maintaining quality. Evaluate its "thinking levels" feature to fine-tune performance for specific tasks like content moderation or UI generation.

Key insights

Gemini 3.1 Flashlight offers high intelligence and speed at a significantly reduced cost for scalable AI workloads.

Principles

Method

The model integrates "thinking levels" in AI Studio and Vertex AI, allowing developers to adjust the model's reasoning depth for specific tasks, optimizing for either speed or complexity.

In practice

Topics

Best for: AI Engineer, CTO, Director of AI/ML, Machine Learning Engineer, AI Architect, MLOps Engineer

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google DeepMind News.