Google ships three new Gemini Flash models but its frontier 3.5 Pro remains lost in training

2026-07-21 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Google has released three new efficient AI models: Gemini 3.6 Flash, 3.5 Flash-Lite, and the specialized 3.5 Flash Cyber, while its anticipated flagship, Gemini 3.5 Pro, remains in private testing. Gemini 3.6 Flash offers 17 percent fewer output tokens and up to 65 percent savings on benchmarks like DeepSWE, priced at \$1.50 per million input and \$7.50 per million output tokens, outperforming 3.1 Pro. Gemini 3.5 Flash-Lite, designed for low latency and high throughput, produces 350 output tokens per second at \$0.30 per million input and \$2.50 per million output tokens. The 3.5 Flash Cyber model, tuned for cybersecurity, scores 83.2 percent on CyberGym, close to OpenAI's GPT-5.5-Cyber (85.6 percent), and identified 55 unique flaws in V8 JavaScript engine commits. Access to 3.5 Flash Cyber is restricted to governments and trusted partners due to its dual-use capabilities. The continued delay of 3.5 Pro leaves Google behind competitors like OpenAI, Anthropic, and Meta in the frontier model race, despite Gemini 4 pretraining being underway.

Key takeaway

For AI Engineers evaluating model deployments, you should consider Google's new Flash models for specific efficiency and cost benefits. Gemini 3.6 Flash offers improved performance over 3.1 Pro at lower prices, while 3.5 Flash-Lite excels in high-throughput scenarios. If your organization handles sensitive code, explore CodeMender with standard Gemini models for automated vulnerability detection. However, be aware that Google's frontier model capabilities currently trail competitors, impacting decisions for cutting-edge applications.

Key insights

Google prioritizes efficiency and specialized AI models while its frontier model lags competitors.

Principles

• Efficiency models can surpass older Pro versions.
• Specialized models offer targeted, high-value capabilities.
• Dual-use AI necessitates restricted access.

In practice

• Use 3.6 Flash for cost-effective, improved agentic tasks.
• Deploy 3.5 Flash-Lite for high-throughput, low-latency workloads.
• Explore CodeMender for automated code security analysis.

Topics

• Gemini 3.6 Flash
• Gemini 3.5 Flash-Lite
• Gemini 3.5 Flash Cyber
• AI Model Efficiency
• Cybersecurity AI
• Frontier Models

Best for: CTO, MLOps Engineer, Investor, AI Engineer, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Google releases three new Gemini models — but no 3.5 Pro — Google prioritizes efficiency and specialized AI models while facing delays in its flagship Pro series amidst intense market competition.
• How Google’s New Gemini Flash Models Compare to its Rivals — Google's new Gemini Flash models deliver competitive speed, cost-efficiency, and specialized capabilities for high-volume AI agent deployment.
• Google releases Gemini 3.1 Flash Lite at 1/8th the cost of Pro — Google's Gemini 3.1 Flash-Lite offers high-speed, cost-efficient AI for scalable enterprise applications.
• Google's Gemini 3.6 Flash model cuts AI agent token costs by up to 65% on long horizon engineering tasks —and 3.5 Pro is on the way — Google's new Flash models prioritize token efficiency and speed for AI agents, significantly reducing operational costs and improving performance.
• 😺 You Need 3 Geminis Now — AI model specialization for specific tasks is replacing the "one model does everything" paradigm.
• Google Introduces Gemini 3.6 to Remind You It Has an AI Model, Too - Gizmodo — Google's new Gemini 3.6 Flash offers token efficiency but lags competitors in benchmarks, while new 3.5 models target cost and cybersecurity.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.