- Published on
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
- Authors
- Name
- DeepHorus
Google DeepMind today released three new proprietary AI models—Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber—aimed at making AI agents faster, smarter, and more cost-efficient at scale. The models promise significant token savings, reducing costs for enterprises on long-horizon engineering tasks by up to 65%. This move underscores Google's push to dominate the rapidly growing market for autonomous AI agents that operate over extended periods.
The new models arrive as Google faces intense competition from rivals like OpenAI and Anthropic, which have released multiple flagship updates since Google's last major model debut. The prior generation Gemini 3.1 Pro launched in February 2026, and developers have been waiting for a more powerful successor. A Google technical staffer confirmed on X that Gemini 3.5 Pro is currently testing with partners and will be released when ready.
Pricing is a key differentiator. Gemini 3.6 Flash costs $1.50 per million input tokens and $7.50 per million output tokens via API. Gemini 3.5 Flash-Lite is even cheaper at $0.30 input and $2.50 output per million tokens. While Google's oldest Gemini 3.1 Flash-Lite remains the most cost-efficient at $0.25/$1.50, the new Flash-Lite is 2x faster, offering better value for speed-sensitive enterprises.
Beyond sticker prices, the models are designed to use fewer tokens overall. In independent benchmarking by the Artificial Analysis Index, Gemini 3.6 Flash reduces output token usage by 17% compared to its predecessor. On the DeepSWE benchmark—measuring multi-step software engineering—token savings reach up to 65%. This efficiency means fewer reasoning steps and tool calls, directly lowering computational costs.
Gemini 3.5 Flash Cyber, a specialized model for cybersecurity researchers and red teamers, has no pricing announced yet. It will be available exclusively to governments and trusted partners via CodeMender, Google's AI code bug-fixing agent. All three models are closed-source, accessible only through Google's official API and partners.
The immediate availability of Gemini 3.6 Flash and Gemini 3.5 Flash-Lite through AI Studio and the consumer Gemini app signals Google's strategic focus on agentic AI. The company is betting that efficiency gains will drive enterprise adoption, even as the broader AI market awaits a true flagship upgrade. With rivals advancing rapidly, Google's latest offerings aim to keep pace by delivering practical, cost-saving solutions today.
