AI Gemini 3.1 Flash-Lite Arrives, Google Promises Fast Performance at a Low Price

Google, via the DeepMind artificial intelligence team, officially introduced the latest AI model named Gemini 3.1 Flash-Lite this week. The model is claimed to be the fastest and most cost-efficient variant in the Gemini 3 Series, while still capable of handling large workloads. One of its key features is support for adjustable thinking-level settings that can be configured on the AI Studio and Vertex AI platforms. Some use cases include: In general, Gemini 3.1 Flash-Lite is described as more capable than Gemini 2.5 Flash, with lower latency, 2.5 times faster in Time to First Answer Token, and up to 45 percent faster output. In performance benchmarks, Gemini 3.1 Flash-Lite recorded an Elo score of 1432 on Arena.ai Leaderboard, a score comparable to OpenAI’s GPT-5.1 or Alibaba’s Qwen3-Max-Preview. The latest Google AI model is also claimed to have higher benchmark scores than its peers in its segment, particularly in the MMMU-Pro reasoning test (76.8 percent) and GPQA Diamond science knowledge (86.9 percent). On pricing, Gemini 3.1 Flash-Lite is pegged at 0.25 USD per 1 million input tokens (characters or one word block) and 1.50 USD per 1 million output tokens (approximately Rp 4,200 and Rp 25,400 respectively). For availability, Gemini 3.1 Flash-Lite is rolling out in preview for developers via Gemini API in Google AI Studio. The model is also available for enterprise users via Vertex AI, as summarised by KompasTekno from Google on Wednesday (4 March 2026). Google cited several companies such as Latitude, Cartwheel, and Whering as starting to use Gemini 3.1 Flash-Lite in early stages to handle large-scale challenges.