Google Introduces Flex and Priority Inference Tiers for Gemini API — 50% Savings or Premium Reliability

Google AI Blog·high signal

Google launched two new inference tiers for the Gemini API: Flex tier offers 50% cost savings ($0.125/M input on Flash-Lite) for latency-tolerant workloads with 1-15 minute target response times using the same synchronous endpoints, and Priority tier offers guaranteed lowest latency at 75-100% over standard rates for business-critical applications. This is Google's answer to the inference cost optimization trend, directly competing with Anthropic's batch API and OpenAI's async endpoints.