Google debuts efficiency-focused Gemini models, delays flagship Pro update

Google DeepMind unveiled a trio of Gemini models this week aimed at helping developers scale AI agents more affordably and quickly. Yet the release highlighted a notable gap: the absence of an updated flagship Pro model, as competitors continue to advance at a rapid clip.

The new models#

The newest lineup includes Gemini 3.6 Flash, designed as the company's core workhorse. This version cuts token consumption by up to 17% compared to its predecessor while bolstering capabilities in coding, information retrieval, and multimodal tasks. Google also introduced Gemini 3.5 Flash-Lite, the most economical option in its class for cost-sensitive deployments.

A third release, Gemini 3.5 Flash Cyber, targets a specialized use case: identifying and remediating cybersecurity flaws. This model will roll out through a controlled access program restricted to government agencies and vetted partners.

The missing piece#

The release underscores a strategic pivot toward efficiency and scale, yet it arrives without updates to Gemini Pro—Google's marquee model for demanding reasoning and coding work. The Pro tier hasn't been refreshed since February, a gap that has widened as the competitive landscape shifted dramatically.

In recent months, OpenAI launched GPT-5.5 and GPT-5.6, while Anthropic rolled out Claude Opus 4.8, Claude Sonnet 5, and expanded access to its frontier Fable 5 model.

Delays and next steps#

In May, Google had signaled an imminent Pro launch, with executives promising a rollout "next month." Last week, Bloomberg reported that internal performance targets were unmet, prompting the delay. On Tuesday, Google DeepMind's Logan Kilpatrick said the company is actively testing Gemini 3.5 Pro with partners and expects to "land soon." He also revealed that work has begun on Gemini 4, the company's most ambitious pre-training effort to date.