Google Introduces Gemini 3.6 Flash and 3.5 Flash-Lite, Says Gemini 4 Is in Training

Google has expanded its artificial intelligence portfolio with the launch of Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, introducing new AI models designed to improve efficiency for developers and enterprises handling high-volume and agentic AI workloads. Alongside the announcements, the company confirmed that Gemini 4 has entered its pre-training phase, marking the beginning of development for its next-generation flagship AI model. The latest offerings are aimed at delivering faster performance, lower operating costs, and enhanced capabilities across coding, reasoning, and multimodal tasks.

According to Google, Gemini 3.6 Flash builds on the capabilities of Gemini 3.5 Flash with notable improvements in software development, document understanding, and complex knowledge-based workflows. The company said the model uses 17 percent fewer output tokens on the Artificial Analysis Index and up to 65 percent fewer tokens on the DeepSWE benchmark by Datacurve, helping reduce computing costs. Benchmark results also indicate stronger performance, with Gemini 3.6 Flash scoring 49 percent on DeepSWE, 63.9 percent on MLE Bench, and 83 percent on OSWorld-Verified, outperforming its predecessor across all three evaluations. Google has priced the model at $1.50 per one million input tokens and $7.50 per one million output tokens.

The newly introduced Gemini 3.5 Flash-Lite has been positioned as the fastest and most affordable model in the Gemini 3.5 family. Google said the model can generate up to 350 output tokens per second, making it suitable for latency-sensitive applications such as agentic search, large-scale document processing, and automated enterprise workflows. Developers can also choose between different reasoning levels, allowing them to balance speed, cost, and computational complexity depending on the workload. Pricing starts at $0.30 per one million input tokens and $2.50 per one million output tokens, making it one of Google's lowest-cost AI offerings to date.

Also Read: Google Alerted Users Ahead Of Venezuela Earthquake Through Android Warning System

Google also highlighted several new enterprise-focused capabilities across its AI ecosystem. Gemini 3.6 Flash includes built-in client-side computer use through the Gemini API and Gemini Enterprise, while supporting multimodal functions such as document parsing, chart analysis, data interpretation, and report generation. The company said the model incorporates enhanced Frontier Safety protections designed to reduce misuse involving chemical, biological, radiological, nuclear (CBRN), and cyber-offensive risks. In addition, Google introduced Gemini 3.5 Flash Cyber, a specialised version fine-tuned to identify, validate, and patch software vulnerabilities in collaboration with Google's CodeMender security agent. Due to its cybersecurity focus, the model will be available only to governments and trusted partners under a limited-access pilot programme.

The company announced that Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available immediately through the Gemini API in Google AI Studio and Android Studio, as well as the Gemini Enterprise Agent Platform. Gemini 3.6 Flash is also accessible through Google Antigravity and the Gemini Enterprise application, while both models are being integrated into the Gemini app. Gemini 3.5 Flash-Lite is additionally rolling out to Google Search. Google also confirmed that Gemini 3.5 Pro remains in partner testing ahead of a broader public release, while revealing that development of Gemini 4 has officially begun with what it described as its most ambitious pre-training run to date. Although no launch timeline has been announced, the disclosure signals Google's continued investment in advancing its generative AI platform amid growing competition in the AI industry.

Also Read: X Android App Receives Design Overhaul, Performance And Reliability Upgrades