Google officially launched Gemini 3.1 Flash-Lite GA on May 7, the fastest and lowest-cost model in the Gemini 3 series, marking Google's latest move in the high-efficiency inference race.
Positioning: The Ultimate Speed-Cost Balance
Flash-Lite is purpose-built for latency-sensitive, high-concurrency enterprise scenarios, covering software engineering, customer service, creative tools, and finance. Google disclosed that the model achieves sub-second response on classification tasks, with p95 latency of about 1.8 seconds under high concurrency, a significant improvement over the previous generation.
Multimodal Capability Landing
Notably, Flash-Lite is the first Lite-tier model in the Gemini 3 series to support multimodality (text + image), along with tool calling and orchestration Agent capabilities, marking that lightweight models can also carry complex Agent workflows.
Pricing: Lowering the LLM Bar Again
Flash-Lite is priced at $0.25 per million input tokens and $1.50 per million output tokens, continuing Google's strategy of continuously compressing cost on high-efficiency models. This is also Google's lowest single-price solution for large-scale enterprise deployment.
Industry Impact
Companies including JetBrains, Gladly, and Ramp have been early adopters in production environments. Google's move to push Flash-Lite to GA is both a response to Preview-stage user feedback and a sign that more high-end models like Gemini 3.2 Flash are about to debut at this year's I/O — the Flash series is becoming Google's anchor point for covering enterprise demand at various price points.