Google expanded its Gemini AI lineup on Wednesday with two powerful new models. The company introduced Gemini 3.6 Flash and Gemini 3.5 Flash-Lite to tackle high-volume AI and agentic tasks more efficiently.
Gemini 3.6 Flash brings meaningful upgrades over its predecessor. Google claims the new model uses 17 percent fewer output tokens on the Artificial Analysis Index. On the DeepSWE benchmark, the reduction reaches an impressive 65 percent.
Developers will also notice faster performance. The model requires fewer reasoning steps and tool calls during complex, multi-step workflows.
Google set competitive pricing for the latest model. Gemini 3.6 Flash costs $1.50 per million input tokens and $7.50 per million output tokens.
Benchmark results show clear progress. On DeepSWE, Gemini 3.6 Flash scored 49 percent compared to 37 percent for Gemini 3.5 Flash. It also achieved 63.9 percent on MLE Bench versus 49.7 percent for the older version. Additionally, it reached 83 percent on OSWorld-Verified.
The model excels at practical work. Users can handle document parsing, chart analysis, data review, and report drafting with ease. It also supports computer use as a built-in tool through the Gemini API.
Google strengthened safety features too. The new model includes enhanced safeguards against chemical, biological, radiological, nuclear, and cyber threats.
At the same time, Google launched Gemini 3.5 Flash-Lite. This version stands out as the fastest and most affordable option in the 3.5 series. It targets high-throughput and low-latency jobs such as agentic search and document processing.
Pricing makes it especially attractive. The model costs $0.30 per million input tokens and $2.50 per million output tokens. It can generate up to 350 output tokens per second.
Developers gain extra flexibility with adjustable thinking levels. They can choose minimal or low thinking for quicker, cheaper responses. Higher levels work better for complicated, multi-step tasks.
Google also revealed a specialized version called Gemini 3.5 Flash Cyber. This model focuses on finding, validating, and fixing software vulnerabilities. It pairs with Google’s CodeMender security agent. However, the company will limit access to governments and trusted partners for now.
Both new models are available immediately. Users can access them through the Gemini API, Google AI Studio, Android Studio, and the Gemini Enterprise platform. They also appear in the Gemini app, while Gemini 3.5 Flash-Lite reaches Google Search as well.
Looking ahead, Google continues pushing boundaries. The company has Gemini 3.5 Pro in testing with partners and expects a wider release soon. Meanwhile, it has started its most ambitious pre-training run yet for the upcoming Gemini 4.
These launches highlight Google’s commitment to making advanced AI faster, smarter, and more accessible for developers and businesses worldwide.
Tech All Over 