Google DeepMind rolled out three new Gemini models this week, and the most telling thing about the release is what’s missing from it. Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber all arrived on schedule. Gemini 3.5 Pro, the flagship reasoning model developers have been asking about since spring, did not.
3.6 Flash
Google’s new workhorse. Sharper at coding, knowledge work, and multimodal tasks, while trimming token consumption by roughly 17% versus 3.5 Flash — making it cheaper to run at scale.
3.5 Flash-Lite
The budget tier of the family, built for developers who need the lowest possible cost per call rather than maximum capability.
3.5 Flash Cyber
A specialist model fine-tuned to find and patch security vulnerabilities. Access is currently restricted to governments and vetted partners through a limited pilot.
Google frames all three as an efficiency play: lower latency and lower cost for the growing wave of customers building AI agents that fire off large volumes of model calls in production. None of the three models is positioned as a reasoning flagship, and Google isn’t pretending otherwise.
That’s the part worth sitting with. The Pro tier is supposed to be Google’s answer to the hardest coding and reasoning workloads — the tier that competes directly with the top-end models from rival labs. It hasn’t been refreshed since February.
The competitive backdrop
The gap is showing up at an inconvenient moment. Since Gemini Pro’s last update, OpenAI has shipped GPT-5.5 and started rolling out GPT-5.6, while Anthropic has released Claude Opus 4.8 and Claude Sonnet 5, and widened access to its frontier Fable 5 model. Every major lab except Google has put out a new top-tier reasoning model in that window.
Why the delay
Reporting from Bloomberg last week pointed to internal delays tied to the model falling short of Google’s own performance targets — not a shelved project, but one that hasn’t cleared the bar internally. Gemini Pro models are meant to be Google’s highest-capability tier for complex reasoning and coding, distinct from Flash models, which are built to be fast and cheap rather than maximally capable.
Google DeepMind product lead Logan Kilpatrick addressed the gap directly this week, saying the company is testing 3.5 Pro with partners now and hopes to land it soon. He also disclosed that the team has kicked off its most ambitious pre-training run yet — for Gemini 4, not 3.5 Pro, suggesting Google may be looking past the current generation rather than rushing it out.
FEB 2026
Gemini Pro last receives a flagship update.
MAY 2026
Google teases 3.5 Pro alongside the 3.5 Flash launch, targeting a rollout “next month.”
JUL 16, 2026
Bloomberg reports internal delays as the model falls short of Google’s performance goals.
JUL 21, 2026
Google ships 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — no Pro. Kilpatrick says testing continues and a Gemini 4 pre-training run has begun.
Timespek’s take
This week’s release is a solid, unglamorous update to the tier of the business that already works: cheaper inference, a dedicated security model, and clear signals that Google is optimizing for agent-scale deployment. What it isn’t is proof that Gemini’s flagship reasoning line is keeping pace. Until 3.5 Pro actually ships, every Flash release doubles as a reminder of the one Google hasn’t delivered — and rivals aren’t waiting around.
