What Google launched on July 21

Google simultaneously released three models, all available immediately on the Gemini API:

  • Gemini 3.6 Flash: the flagship model of the Flash line, optimized for response speed without sacrificing too much reasoning.
  • Gemini 3.5 Flash-Lite: an even lighter variant, designed to push cost per call to the minimum on high-volume tasks.
  • Gemini 3.5 Flash Cyber: a variant specialized in cybersecurity tasks — log analysis, intrusion pattern detection, alert classification.

What stands out is what did not ship with them: Gemini 3.5 Pro, the generation's most capable model, is still in testing with no confirmed public general availability date.

Why Google prioritizes Flash over Pro

The strategy makes business sense: the vast majority of API call volume in production —classification, data extraction, short answers, content moderation— does not need the top-of-the-line reasoning of a Pro model. It needs to be fast, cheap and reliable at volume.

The industry pattern: Google is not the only example — Anthropic, OpenAI and Z.AI have also been iterating faster on their budget tiers than on their flagship models. The real competition today is as much about the capability ceiling as about the cost-per-token floor.

Flash Cyber: the first Gemini model specialized in security

Gemini 3.5 Flash Cyber is the first variant of the Flash family with explicit vertical specialization. It is tuned for cybersecurity tasks: security log parsing, event correlation, detection rule generation, and alert triage — a market where Google competes directly with Microsoft Security Copilot and specialized SOC tools.

This vertical specialization marks a trend to watch: the big labs are starting to release variants tuned by industry, not just by size/speed, betting that domain specificity makes up for what is lost in generality.

What it means for those building on the Gemini API

If your product uses Gemini for volume tasks (classification, extraction, short summaries), it is worth evaluating a migration to 3.6 Flash or 3.5 Flash-Lite as soon as pricing is confirmed — historically each Google Flash generation improves quality at the same or lower cost than the previous one.

If your use case depends on deep reasoning (complex legal analysis, advanced math, highly difficult code), the recommendation is to wait for the general availability of 3.5 Pro before migrating — using Flash for those tasks usually produces a noticeable drop in quality.