Google releases three new Gemini models — but no 3.5 Pro
Original reporting by TechCrunch
Google DeepMind's Gemini Flash models are a family of AI models optimized for efficiency, speed, and cost-effectiveness, intended for high-volume production applications. On Tuesday, Google DeepMind unveiled Gemini 3.6 Flash, alongside 3.5 Flash-Lite and 3.5 Flash Cyber, emphasizing a commitment to delivering superior efficiency, lower latency, and enhanced reliability for customers developing AI agents at scale. The new 3.6 Flash model serves as Google’s “workhorse,” promising improved capabilities across coding, knowledge work, and multimodal performance, while reducing token usage by up to 17% compared to its predecessor, making it more economical. The 3.5 Flash-Lite offers the lowest cost, and the specialized 3.5 Flash Cyber is fine-tuned for identifying and remediating cybersecurity vulnerabilities, available through a limited access pilot for governments and trusted partners.
The Pro omission
This release is notable not just for the advanced, more affordable, and specialized models shipped, but also for what was conspicuously absent: an update to Google's flagship Gemini Pro model. Last updated in February, the Pro series typically handles more complex reasoning tasks. Its continued absence comes amidst a rapid succession of frontier model releases from rivals like OpenAI and Anthropic. Google had previously teased a 3.5 Pro update for May, but recent reports suggest internal delays as the company reportedly struggles to meet its own performance benchmarks. While Google DeepMind states that Gemini 3.5 Pro is currently being tested with partners and is expected "soon," the focus on efficiency with Flash models marks a strategic, if perhaps reactive, pivot in its immediate offering.
The rollout of Gemini 3.6 Flash and its counterparts underscores Google DeepMind’s commitment to providing developers with efficient, cost-effective AI tools for building at scale. These models, optimized for tasks ranging from coding to specialized cybersecurity applications, represent a practical, immediate solution for many enterprise needs, prioritizing speed and affordability over raw frontier capabilities. However, the release is conspicuously marked by the continued absence of Gemini 3.5 Pro, a highly anticipated flagship model designed for complex reasoning, which was previously expected by May. This divergence highlights a strategic tension between delivering readily available, production-grade solutions and the challenging pursuit of cutting-edge frontier AI.
The Shifting Landscape
This delay, amidst a flurry of advanced model releases from competitors like OpenAI and Anthropic, signals potential internal hurdles for Google in the intense race for AI leadership. While Google hints at Gemini 3.5 Pro's imminent arrival and the ambitious pre-training for Gemini 4, the current situation might prompt some developers to look to rival offerings for immediate, cutting-edge reasoning tasks. The broader implication suggests a rapidly evolving market where practical application models are becoming commoditized, even as the top-tier battle for AI supremacy remains fiercely contested. Google's current emphasis on Flash models indicates a strong focus on scalable agentic AI, where cost and latency are critical. Yet, the delayed Pro release underscores the profound technical challenges of pushing the boundaries of AI, leaving Google under persistent pressure to deliver its highest-capability models and maintain its position at the forefront of innovation.
Frequently asked questions
- What new AI models did Google DeepMind recently release for developers and businesses?
- Google DeepMind launched Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. Gemini 3.6 Flash is a versatile workhorse model offering improved coding and multimodal capabilities with reduced costs. 3.5 Flash-Lite prioritizes cost-effectiveness, while 3.5 Flash Cyber is specialized for finding cybersecurity vulnerabilities, available to governments and trusted partners. These models focus on efficiency and reliability for building AI agents at scale.
- Why did Google DeepMind delay the expected release of its Gemini Pro model update?
- Google DeepMind's anticipated Gemini Pro update was delayed due to internal struggles to meet performance goals. The company had initially hinted at a May release but faced challenges in development. While the Flash models prioritize speed and cost-efficiency, Gemini Pro models are designed for complex reasoning and higher capabilities, requiring rigorous internal testing before public rollout.
- What are the key differences between Google's Gemini Flash and Gemini Pro AI models?
- Google's Gemini Flash models are optimized for lower cost and faster response times, making them ideal for high-volume production applications and efficiency. In contrast, Gemini Pro models represent Google's highest-capability offerings, designed for more complex reasoning tasks and advanced coding. While Flash prioritizes scale and affordability, Pro focuses on intricate problem-solving and top-tier performance, often for more demanding applications.