Google Unveils Gemini 3.5 Flash with Focus on Speed and Triple the Price

On Tuesday, May 19, 2026 (UTC), tech giant Google had confidential details of its new AI model leaked just hours before the opening of the annual Google I/O 2026 event in Mountain View. Independent developer pankajkumar_dev revealed that the company will launch the Gemini 3.5 Flash (internally identified as gemini-3.5-flash), a model optimized for real-time production that prioritizes execution speed over pure cost reduction.

Price and Performance Trade-off

Contrary to the recent industry trend of lowering token costs, Google's new bet increases computation values to deliver record execution time responses. According to the pricing tables published by developer ayushrajgorar, the input cost per million tokens rose to $1.50, a threefold increase from the traditional Gemini 3 Flash, which costs $0.50. The output rate is set at $9.00 per million tokens, compared to $3.00 for the previous generation model.

In practice, the market gains alternatives for different business needs. For systems running in the background that can tolerate delays, the Flex tier reduces input costs to $0.75. On the other hand, for industrial applications relying on instant responses, the Priority tier charges $2.70 per million tokens to ensure the lowest possible processing queue time.

Optimized Infrastructure and Sub-200ms Latency

Internal console tests show that the new model achieves response latency below 200 milliseconds in standardized production queries. This technical advancement was achieved through a combination of robust distillation of larger models and sparse hardware architectures. The model also features logical reasoning capabilities similar to the Gemini 3.1 Pro, and includes enhanced grounding and verification systems to significantly reduce the occurrence of inaccurate or hallucinated responses.

Many programmers in the community are debating on social media platform X whether the higher cost will be offset by operational stability. The expectation is that the official announcement and the opening of public API keys will occur during the main presentation of Google I/O, scheduled for today at 17:00 (UTC) on the official site io.google.