Google releases three new Gemini models — but no 3.5 Pro

Google's Strategic Model Expansion

Google DeepMind has unveiled three new additions to its Gemini model family: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. This release underscores Google's commitment to providing more efficient, cost-effective, and specialized AI capabilities for developers building at scale. Gemini 3.6 Flash, positioned as Google's “workhorse model,” boasts enhanced performance in coding, knowledge work, and multimodal tasks, while significantly reducing token usage by up to 17% compared to its predecessor, 3.5 Flash, making it a more economical choice.

The new models are specifically designed to address the growing demand for “efficiency, latency, and reliability” in large-scale AI agent development. Gemini 3.5 Flash-Lite stands out as the most cost-effective option within this class, catering to applications where budget is a primary concern. Furthermore, the introduction of 3.5 Flash Cyber marks a strategic move into specialized AI, fine-tuned for identifying and rectifying cybersecurity vulnerabilities. This particular model will be exclusively accessible to governments and trusted partners through a limited access pilot program, highlighting its sensitive and critical application. These Flash models are distinct from the Pro versions, which typically offer higher capabilities for complex reasoning but at a potentially greater cost and latency.

The Missing Flagship and Competitive Pressures

Despite the new releases, the announcement was notably marked by the absence of the highly anticipated Gemini 3.5 Pro, Google's flagship model. This omission is significant, especially given that Google had previously teased its release in May, stating it was “already being used internally, and we look forward to rolling it out next month.” The delay appears to stem from internal challenges, with Bloomberg reporting that Google has been struggling to meet its own internal performance goals for the 3.5 Pro model.

This setback places Google in a challenging position within the fiercely competitive AI landscape. Rival labs have maintained an aggressive release pace, with OpenAI having already launched GPT-5.5 and begun rolling out GPT-5.6. Similarly, Anthropic has advanced its offerings with Claude Opus 4.8 and Claude Sonnet 5, alongside expanding access to its frontier Fable 5 model. The continuous stream of updates from competitors underscores the intense pressure on Google to deliver its most capable models, particularly for complex reasoning and coding tasks where Gemini Pro is expected to excel. The delay could potentially allow competitors to further entrench their lead in the frontier AI space.

Looking Ahead: Pro and Gemini 4

Despite the current delay, Google DeepMind product lead Logan Kilpatrick offered an update on the status of Gemini 3.5 Pro, stating that the company is actively testing it with partners and anticipates its release “soon.” This suggests that while internal hurdles exist, Google remains committed to bringing its flagship model to market. The testing phase with partners indicates a focus on ensuring the model meets external performance expectations before a broader rollout.

Looking further into the future, Kilpatrick also revealed that the Google DeepMind team has initiated its “most ambitious pre-training run yet for Gemini 4.” This forward-looking statement signals Google's long-term vision and continued investment in developing next-generation AI capabilities, even as it navigates the immediate challenges with its current Pro model. The ambition behind Gemini 4 suggests Google is aiming for a significant leap in AI performance, potentially addressing the competitive gap observed with the 3.5 Pro delays.