Alongside a cyber model: Google presents the next generation of Gemini
Google has announced Gemini 3.6 Flash and additional dedicated models to enable the deployment of AI agents at scale. Alongside significant cost savings and exceptional output rates, the company confirms that pre-training of Gemini 4 has already begun.
The technology giant Google has officially unveiled a major upgrade to its Flash model series, designed to address one of the primary challenges for developers today: the cost, latency, and resource consumption required to run artificial intelligence agents in production.
At the center of the announcement is Gemini 3.6 Flash, the core model that demonstrates a significant improvement in coding performance, knowledge tasks, and multimodal processing, alongside a significant decrease in token consumption. According to the Artificial Analysis index, the new model shows savings of about 17% in output tokens compared to the 3.5 Flash model, and in certain performance tests (such as DeepSWE), a reduction of up to 65% in the tokens required to perform the task was recorded.
This efficiency, coupled with a price of $1.50 per million input tokens and $7.50 per million output tokens, is intended to dramatically lower the cost of continuous operation for autonomous agents.
Record speed and dedicated cybersecurity
Alongside the main Flash model, Google launched two additional focused models: Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber.
-
Gemini 3.5 Flash-Lite: The fastest model in the series, reaching a processing rate of 350 tokens per second. At an accessible price of $0.30 per million input tokens and $2.50 per million output tokens, the model is intended for high-volume tasks such as real-time document processing, receipt analysis, and data scanning.
-
Gemini 3.5 Flash Cyber: A model that has undergone fine-tuning dedicated to identifying, verifying, and fixing security vulnerabilities in code. The model is integrated into the CodeMender code security platform and will be available in the first stage as part of a limited pilot for governments and authorized security agencies only.
Looking toward Gemini 4
In parallel with the launch of the new models, Google notes that the Gemini 3.5 Pro model is currently in advanced testing stages with business partners and is expected to be released to the general public as soon as it is ready.
Additionally, the company provided a glimpse into its future roadmap, reminding the industry that the technological race is far from over: Google's development teams have officially begun the pre-training of Gemini 4, which is defined by the company as the most extensive and ambitious language model project in its history.
Availability
Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available for developers and enterprise customers in Gemini Enterprise. The 3.6 Flash model is now also available to all users of the Gemini app, while the Flash-Lite model is being gradually integrated into the company's search engine.
