Google has expanded its Gemini artificial intelligence portfolio with the launch of three new models designed for developers, enterprises, and cybersecurity applications, even as its highly anticipated flagship model, Gemini 3.5 Pro, remains delayed for the third time.

Announced on Tuesday, the new releases include Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber—a specialized cybersecurity model available only to governments and select security partners.

The launch underscores Google’s strategy of strengthening its AI ecosystem through faster, more efficient models while work continues on its next-generation flagship system.

Focus on Speed, Cost and Efficiency

The latest additions are aimed at improving coding performance, accelerating AI agents, and reducing computing costs for businesses deploying large-scale AI applications.

According to Google, Gemini 3.6 Flash delivers stronger performance than its predecessor while lowering both the number of tokens required to generate responses and overall output costs.

The company says the model is optimized for software development, enterprise productivity, and knowledge-intensive tasks, allowing organizations to process larger workloads more efficiently.

Google has increasingly emphasized affordability as a competitive advantage. Chief Executive Officer Sundar Pichai has previously argued that deploying Flash-class models at scale could save businesses more than $1 billion annually through improved efficiency and lower inference costs.

Improved Coding Performance

Google also highlighted significant gains across several technical benchmarks.

On one software engineering evaluation, Gemini 3.6 Flash improved its score from 37 percent to 49 percent, while its performance on a machine learning research benchmark increased from 49.7 percent to 63.9 percent.

The model also recorded higher scores on computer operation tasks, improving from 78.4 percent to 83.0 percent, reflecting Google’s continued focus on practical coding and automation capabilities.

The company hopes the improvements will encourage developers to continue building AI-powered applications on Google’s platform while more advanced systems remain under development.

Gemini 3.5 Pro Remains Delayed

Despite the new releases, attention remains firmly focused on the absence of Gemini 3.5 Pro, Google’s next flagship AI model.

Originally expected to debut following announcements made during Google I/O, the model has now missed its third scheduled launch date, including its latest target of July 17.

Reports indicate that Google abandoned an earlier version of the model and restarted development from the pretraining stage after engineers encountered performance challenges.

The setbacks have reportedly been linked to persistent hallucination issues, reliability concerns, and coding performance that failed to meet Google’s internal quality benchmarks.

Industry reports have also suggested that several researchers involved in the project have departed during the development process, adding to speculation about the challenges facing the company’s flagship AI efforts.

Competitive Pressure Intensifies

The delays come at a particularly competitive moment for Google.

Over the past week, several major AI companies—including OpenAI, Anthropic, Meta, xAI, and Moonshot AI—have introduced new frontier models, intensifying competition in the rapidly evolving artificial intelligence sector.

Industry analysts also note that Google recently dropped out of the top five on a closely watched AI capability ranking for the first time, highlighting how quickly the competitive landscape has shifted.

The increasingly crowded market has placed additional pressure on Google to deliver a flagship model capable of competing with the industry’s leading AI systems.

Looking Ahead

Even as development continues on Gemini 3.5 Pro, Google has confirmed that work has already begun on Gemini 4, signaling that the company is pursuing multiple long-term AI development tracks simultaneously.

For now, Google’s strategy appears focused on maintaining momentum through incremental improvements to its Flash model family while engineers continue refining the delayed flagship system.

Analysts say the approach reflects the changing dynamics of the AI industry, where rapid iteration, lower operating costs, and reliable deployment are becoming just as important as launching the most powerful model.

Whether the interim releases will be enough to satisfy developers—and help Google regain leadership in the increasingly competitive AI race—may ultimately depend on when Gemini 3.5 Pro is finally ready for release.