OutYet reporting
Google's Gemini Flash split sharpens the fast-model decision
Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber separate general agent work, throughput-sensitive workloads, and cyber-agent use cases more clearly than a single Flash label did.
Google's July 21 announcement groups three models under a new Gemini Flash lineup: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. Google positions 3.6 Flash as the general workhorse for coding, knowledge work, and multimodal tasks; it positions Flash-Lite for high-throughput and low-latency work; and it presents Flash Cyber alongside the CodeMender security agent. The important product signal is specialization within the fast-model tier, not a single universal replacement choice.
For 3.6 Flash, Google reports 17% fewer output tokens than Gemini 3.5 Flash on the Artificial Analysis Index and lists prices of $1.50 per million input tokens and $7.50 per million output tokens. It also reports higher scores than 3.5 Flash on DeepSWE, MLE Bench, and OSWorld-Verified. Those figures make the claimed improvement about task efficiency as well as benchmark score, but they remain Google's reported results and should not be read as a universal cost reduction.
Flash-Lite is the more distinct operational option. Google calls it its fastest 3.5-series model, cites 350 output tokens per second from Artificial Analysis, and lists $0.30 per million input tokens and $2.50 per million output tokens. Google also says it supports configurable thinking levels and built-in computer use. Flash Cyber has a narrower role: Google describes it as a cyber-focused model paired with CodeMender, rather than a general model choice for every coding or agent workload.
Technical users should therefore select by workload shape before treating the family as a leaderboard contest. A high-volume extraction or routing service may benefit from the throughput and pricing profile Google describes for Flash-Lite, while a mixed coding and document workflow is the stated target for 3.6 Flash. The announcement cites external benchmarks but does not provide an independent production comparison for a specific application, so evaluation should include tool reliability, output length, safety behavior, and total task cost on representative traces.