Gemini 3 Flash Preview (reasoning)

Gemini 3.1 Pro Preview is Google's most advanced reasoning Gemini model, designed to solve complex problems. It improves upon the performance and reliability of the Gemini 3 Pro line, delivering better thinking capabilities, enhanced token efficiency, and a more practical and consistent experience. This model can understand large datasets and difficult problems from diverse sources of information, including text, audio, images, video, PDFs, and even entire code repositories.

Gemini 3.5 Flash is Google DeepMind's latest efficiency-focused language model, announced at Google I/O 2026. It is the next iteration in the Gemini 3 series, outperforming Gemini 3.1 Pro on coding and agentic tasks. Gemini 3.5 Flash delivers impressive speed, though it comes at a relatively high price point — $1.5 per 1M input tokens and $9 per 1M output tokens.

Gemini 3 Flash Preview (non-reasoning) is one of the leading models in terms of intelligence and offers good pricing compared to non-reasoning models in the same price range. This model is also noticeably fast, though slightly verbose. It supports text, image, audio, and video input, outputs text, and has a 1 million token context window with knowledge updated to January 2025.

Gemini 3 Pro Preview (high) is one of the leading models in intelligence, but has a relatively high cost compared to models in the same price range. This model is also very fast but tends to be quite verbose. It supports text, image, audio, and video input, while outputting text.

Gemini 3 Pro Preview (low) is the speed- and cost-optimized variant of the premium Gemini 3 Pro model during the Preview phase. It is designed to handle everyday tasks and high-volume workloads, prioritizing fast response times and low latency over deep reasoning. However, this version still maintains core multimodal capabilities, able to efficiently process and understand data from text, images, and audio. It is an ideal choice for applications requiring rapid interaction and cost efficiency.