On July 21, 2026, Google unveiled three new Gemini AI models. These models focus on efficiency and cost savings. At the same time, Google delayed its flagship Gemini 3.5 Pro model.
The new releases include Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. Each serves a different purpose, aimed at different users and tasks.
Gemini 3.6 Flash: The Workhorse Model
Google calls Gemini 3.6 Flash its “workhorse model.” It is designed for coding, knowledge work, and multimodal tasks. This model processes tasks faster and uses fewer tokens than its predecessor.
Specifically, 3.6 Flash uses about 17% fewer tokens than Gemini 3.5 Flash. That means it costs less to run. The pricing is $1.50 per million input tokens and $7.50 per million output tokens.
In coding tests like DeepSWE, 3.6 Flash shows 49% better performance than 3.5 Flash. It can also handle computer use as a standard feature in the Gemini API. This makes it a strong choice for developers and knowledge workers.
Gemini 3.5 Flash-Lite: Speed and Low Cost for Volume
Flash-Lite targets high-volume automation and document processing. It offers lower rates, making it cheaper for heavy use. The cost is $0.30 per million input tokens and $2.50 for output tokens.
This model can output 350 tokens every second, ideal for tasks where speed and volume matter more than complexity. Google expects to see a lot of Flash-Lite in Google Search, where fast AI responses are key.
Gemini 3.5 Flash Cyber: Focused on Security
Flash Cyber is a limited pilot version. It is fine-tuned for cybersecurity tasks, especially finding vulnerabilities. This model is only available to governments and trusted partners through a program called CodeMender.
It’s not a wide release but shows Google’s focus on AI for security. Gemini 3.5 Pro, the flagship cybersecurity model, remains in testing with unnamed partners. It was expected to launch in June but was delayed.
Google claims the 3.5 Pro is almost as good at finding and fixing cybersecurity issues as Claude Mythos, a competitor’s model. But it’s not publicly available yet.
Looking Ahead: Gemini 4 and the Future
While the new models arrive, Google has already started pre-training Gemini 4. This next flagship model is expected to be more ambitious than past versions.
There is no timeline for Gemini 4’s public release, but it won’t come for several months. The delay leaves room for competitors like OpenAI and Anthropic, who have also launched new models recently.
Google made changes to 3.6 Flash based on user feedback from the 3.5 Flash release. This shows Google listens and improves its models regularly.
For now, Gemini 3.6 Flash is the best choice for most users needing a powerful, versatile AI. Flash-Lite suits high-speed, high-volume needs. Flash Cyber is reserved for security-focused partners.
The shift from 3.5 Flash to 3.6 Flash means users get better performance and lower costs. Google’s AI lineup is growing, but the flagship Gemini 3.5 Pro remains a work in progress.
Based on
- Which New Gemini Model Should You Use: 3.6 Flash, Flash-Lite or Cyber? — justainews.com
- Google Doubles Down on Faster, Cheaper AI — but No Sign of Gemini 3.5 Pro – Business Insider — businessinsider.com
- Google releases three new Gemini models — but no 3.5 Pro | TechCrunch — techcrunch.com
- Google announces Gemini 3.6 Flash and cybersecurity AI, teases 3.5 Pro and Gemini 4 – Ars Technica — arstechnica.com
- Rivals Poke Fun at Google’s Delayed Gemini 3.5 Pro Model – Business Insider — businessinsider.com
