Last updated: August 22, 2026 — verified against Google’s current Gemini API model catalog, release notes and pricing documentation.
Google’s Gemini family now includes several specialized model lines rather than one simple Pro-versus-Flash split. The newest general-purpose Flash release is Gemini 3.7 Flash, while Gemini 3.5 Flash-Lite targets high-volume low-cost workloads and Gemini 3.1 Pro Preview remains available for Pro-tier reasoning use cases.
Gemini 3.7 Flash
Gemini 3.7 Flash became generally available on August 13, 2026. Google describes it as its most capable Flash model for complex coding, agentic workflows and reliable multi-step execution. It supports a 1M-token context window, up to 64k output tokens, multimodal input, thinking levels, function calling, code execution, search grounding and preview computer use.
Gemini 3.5 Flash-Lite
Gemini 3.5 Flash-Lite is Google’s most cost-efficient GA model for high-volume agentic tasks, translation and simpler data processing. Its standard paid pricing is $0.30 per million input tokens and $2.50 per million output tokens.
Gemini 3.1 Pro Preview
Gemini 3.1 Pro Preview remains available in the API for more demanding reasoning and multimodal workloads. However, Google’s current migration guidance now directs many older Flash and Pro workloads toward Gemini 3.7 Flash, so new deployments should benchmark the latest Flash model rather than assuming Pro is automatically the best choice.
Image and Specialized Gemini Models
The Gemini catalog also includes image-focused models such as Nano Banana 2 and Nano Banana 2 Lite, along with specialized audio, robotics and research endpoints. These should be treated as separate task-specific model families rather than part of a single text-model ranking.
How to Choose a Gemini Model
- Complex coding and agents: start with Gemini 3.7 Flash.
- High-volume low-cost execution: use Gemini 3.5 Flash-Lite.
- Pro-tier reasoning requirements: benchmark Gemini 3.1 Pro Preview against 3.7 Flash on your exact workload.
- Image generation/editing: use the dedicated Nano Banana family rather than a general text model.
For a deeper side-by-side comparison with pricing, see our Gemini Model Comparison.
