Google introduced Gemini 3.5 Flash, a new AI model with frontier-level intelligence designed for agentic workflows and coding, available worldwide starting May 19, 2026 [1, 2, 3, 4, 5, 6, 7, 8]. The model delivers up to four times faster output tokens per second than competing frontier models, running at around 280 tokens per second to enable efficient long-horizon autonomous tasks [1, 2, 3, 4, 6, 8]. Gemini 3.5 Flash operates at less than half the cost of similar models, sometimes as low as a third, making it a cost-effective solution for developers and enterprises [1, 4, 6].
The new model outperforms Google's previous flagship Gemini 3.1 Pro on multiple benchmarks. It scored 76.2% on Terminal-Bench 2.1, earned a 1656 Elo rating on GDPval-AA, achieved 83.6% on MCP Atlas, and reached 84.2% on CharXiv Reasoning tests [1, 3, 6, 8]. DeepMind's chief technologist Koray Kavukcuoglu said, "3.5 Flash offers an incredible combination of quality and low latency. It outperforms our latest frontier model, 3.1 Pro, on nearly all the benchmarks" [3].
Gemini 3.5 Flash powers multiple Google products, including the Gemini app, Google Search AI Mode, Google AI Studio, Android Studio, and the Gemini Enterprise Agent Platform for businesses [1, 3, 6, 7, 8]. It also underpins Gemini Spark, a personal AI agent that runs continuously on Google Cloud without requiring user hardware, offering persistent AI assistance [9, 7].
Google's new AI agent platform "Antigravity" is optimized for Gemini 3.5 Flash and supports collaborative multi-agent workflows and coding pipelines. Google Senior Director Tulsee Doshi highlighted, "With post-training, we’re really starting to unlock some of the value of the feedback we’re getting from users, for example, from Antigravity. That’s really what you’re seeing play out in terms of the code performance and the tool use performance" [2].
At Google I/O 2026, CEO Sundar Pichai emphasized the focus on practical AI applications and stated, "Gemini 3.5 Flash is a first in a series of models combining frontier intelligence with actions" [4, 6]. On Gemini 3.5 Pro, a more powerful model currently used internally, Pichai said, "I know you can't wait to get your hands on it. Give us until next month to get [Gemini 3.5 Pro] to you" [10].
Alongside Gemini 3.5 Flash, Google announced Gemini Omni Flash, a multimodal video generation and editing model available to Google AI Plus, Pro, and Ultra subscribers, plus YouTube content creators [4, 6].
Google plans to release Gemini 3.5 Pro in June 2026 after delaying its launch to improve AI coding task performance, using feedback from the widely deployed 3.5 Flash model [10, 11].
