Google ships cheaper Flash updates but its Gemini Pro model remains MIA

Google DeepMind dropped a trio of new models on Tuesday — but not the one developers have been waiting for. The headliner is Gemini 3.6 Flash, billed as an improved “workhorse” for coding and multimodal tasks that also trims token costs by up to 17% compared to its predecessor. Alongside it, the company launched 3.5 Flash-Lite, which it calls the most cost-effective model in its class, and 3.5 Flash Cyber, a specialized variant fine-tuned for finding and patching cybersecurity vulnerabilities.

That last one comes with a catch. Flash Cyber won’t appear in public APIs anytime soon. Google is restricting access to governments and vetted partners through a limited pilot, a nod to the sensitive nature of automated vulnerability detection and the geopolitical climate around AI security tools.

The practical focus here is clear. Product lead Logan Kilpatrick framed the releases around efficiency, latency, and reliability — the unglamorous but essential ingredients for companies building AI agents at scale. A 17% cost reduction on a high-volume workhorse model isn’t flashy, but for startups running millions of inference calls, it changes the math.

What’s missing, though, is the elephant in the room. Gemini 3.5 Pro, teased months ago as the next flagship reasoning model, is still nowhere to be found. Google said in May that Pro was in internal use and would ship the following month. It didn’t. Bloomberg reported last week that the model is facing delays because it’s not meeting internal performance benchmarks. Meanwhile, OpenAI has shipped GPT-5.5 and begun rolling out GPT-5.6. Anthropic launched Claude Opus 4.8 and Sonnet 5. Google’s Flash updates keep the lights on for enterprise customers, but the top-tier race isn’t waiting.