The latest models from the Gemini team aim to provide higher efficiency, lower latency, and enhanced reliability for building AI agents at scale.
3.6 Flash
Gemini 3.6 Flash serves as our workhorse model, delivering improved coding, knowledge work, and multimodal performance. According to the Artificial Analysis Index, it reduces output token usage by 17% compared to 3.5 Flash, with reductions of up to 65% in benchmarks like DeepSWE by Datacurve, all while lowering the cost per output token.
Priced at $1.50 per million input tokens and $7.50 per million output tokens, 3.6 Flash reduces the overall cost for building and running agents. It shows higher precision across use cases, with fewer unwanted code edits and reduced execution loops, achieving 49% accuracy in DeepSWE compared to 37% for 3.5 Flash.
3.5 Flash-Lite
3.5 Flash-Lite is our fastest and most cost-effective model in the 3.5 series, operating at 350 output tokens per second, significantly enhancing developer workflows. Priced at $0.3 per million input tokens and $2.5 per million output tokens, it offers a strong price-performance ratio.
3.5 Flash Cyber
3.5 Flash Cyber is focused on cybersecurity applications, combining an efficient model with the CodeMender code security agent to effectively detect and patch vulnerabilities. The model will be available exclusively to governments and trusted partners to ensure an edge in frontline defenses.
Conclusion
3.6 Flash and 3.5 Flash-Lite are available now, and we welcome developers to start using them via Google AI Studio and Android Studio, looking forward to feedback for future improvements.
Blogger's Review: The latest release of the Gemini series undoubtedly provides robust technical support for building AI agents, especially with significant advancements in efficiency and cost control, which will further drive AI applications across various industries.