Google announced Gemini 3.1 Flash‑Lite as the “fastest and most energy‑efficient member of its lineup.”

Google introduces Gemini 3.1 Flash‑Lite – an economical and fast AI model

Google announced the launch of a new version of the Gemini 3.1 Flash‑Lite model, positioned as the most affordable in the Gemini lineup in terms of cost and performance.

* Big data optimization – the algorithm is specifically tuned for efficient processing of large data sets.

* Usage cost – $0.25 per million input tokens and $1.50 per million output tokens.

A preview version is already available:

- to developers via the Gemini API in Google AI Studio;
- to corporate clients – in Vertex AI.

Speed and quality
Metric | Gemini 3.1 Flash‑Lite | Gemini 2.5 Flash
---|---|---
Time to first token | 2.5× faster |
Output speed | 45 % higher |
Answer quality | comparable/higher |

Low latency makes the model especially attractive for real‑time applications: from chatbots to interactive interfaces where instant response is required.

Test results
* ELO rating on Arena.ai – 1432 points.
* GPQA Diamond – 86.9 %.
* MMMU Pro – 76.8 %.

In these tests Gemini 3.1 Flash‑Lite outperformed models of similar level and even some larger predecessors (e.g., Gemini 2.5 Flash) in reasoning and multimodal processing.

Flexibility and scalability
Developers can adjust the depth of the model’s reasoning to fit their tasks – an important factor for high‑frequency workloads.

The algorithm is optimized for:

- translating large volumes of text;
- content moderation where cost is paramount.

It also suits UI generation, dashboards, modeling, and research based on complex queries.

In short: Gemini 3.1 Flash‑Lite combines affordability, speed, and reliable quality, making it a valuable tool for both individual developers and corporate clients seeking low‑latency, high‑performance AI solutions.