Google released three new Gemini AI models on July 21, bringing faster inference and lower pricing to its flagship large language model lineup.

The new models—Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber—mark Google’s ongoing push to make advanced AI accessible at scale while tightening its grip on the inference cost war.

Pricing and Performance Gains

Gemini 3.6 Flash costs $1.50 per million input tokens and $7.50 per million output tokens, down from the previous Flash tier’s $9 per million output pricing. That’s a cut of roughly 17% on output costs.

The performance gains match the cost reduction. Gemini 3.6 Flash uses about 17 percent fewer output tokens on the Artificial Analysis Index benchmark compared to earlier models and completes multi-step tasks with fewer reasoning steps and tool calls.

The new Flash-Lite tier delivers even cheaper inference for simpler workloads at $0.30 per million input tokens and $2.50 per million output tokens.

Security-Focused Variant Launches

Google also introduced Gemini 3.5 Flash Cyber, a security-hardened variant restricted to governments and trusted partners. The model is built for threat assessment and cybersecurity analysis in trusted environments only.

Conspicuously absent from the lineup was Gemini 3.5 Pro, Google’s flagship model. The release has now missed its target timeline multiple times. Google confirmed it has begun “our most ambitious pre-training run yet” for Gemini 4, suggesting a longer development cycle.

What This Means

Faster, cheaper inference moves the value proposition of AI applications downstream to developers and enterprises. With lower per-token costs, use cases that were unprofitable at previous pricing—like real-time summarization, personalized search, and agentic workflows—become economically viable.

The missing Pro model signals Google may be rethinking its product tier strategy. Gemini 4 could represent a larger architectural jump than incremental Flash improvements.

Google is betting speed and cost will matter more than raw capability as AI commoditizes.