Google's Triple Gemini Launch: 3.6 Flash, Flash-Lite, Cyber & Gemini 4 Pre-training Begins

Google DeepMind announced three new Gemini models in a single release, each targeting distinct production workloads, while simultaneously confirming that pre-training for the next-generation Gemini 4 has begun.

Gemini 3.6 Flash: Token-Efficient Main Model

Positioned as the flagship of the trio, Gemini 3.6 Flash emphasizes reduced token consumption without sacrificing capability. According to Artificial Analysis Index benchmarks, it uses 17% fewer output tokens than its predecessor, 3.5 Flash. On the DeepSWE coding benchmark, token savings reach up to 65%. The model also requires fewer reasoning steps and tool calls to complete multi-step tasks, translating to lower latency and cost.

Pricing is set at $1.50 per million input tokens and $7.50 per million output tokens, undercutting 3.5 Flash. Despite the lower cost, 3.6 Flash improves on key benchmarks:

DeepSWE (coding): 37% → 49%

MLE Bench (ML research): 49.7% → 63.9%

OSWorld-Verified (computer control): 78.4% → 83%

GDPVal-AA v2 (knowledge work): 1,421 tasks completed, ~70 more than previous generation

Demonstrations show 3.6 Flash orchestrating multi-agent workflows for complex code migration, generating 3D texture-extraction tools via Gemini Canvas, and acting as an AI interaction designer for immersive theme studios.

Gemini 3.5 Flash-Lite: Speed and Volume Optimized

Flash-Lite targets high-throughput, cost-sensitive scenarios such as bulk document processing and agent-driven search. It delivers 350 output tokens per second — the fastest in the 3.5 series — at $0.30 per million input tokens and $2.50 per million output tokens.

Despite its lightweight profile, Flash-Lite outperforms the larger Gemini 3 Flash on several benchmarks:

SWE-Bench Pro: 54.2% vs. 49.6%

OSWorld-Verified: 74.0% vs. 65.1%

The suggested architecture pairs 3.6 Flash as the "planner" decomposing tasks with multiple Flash-Lite instances as parallel "workers." A demo generated 25 production-ready web designs in rapid succession. Flash-Lite is already live in the Gemini App and slated for integration into Google Search.

Gemini 3.5 Flash Cyber: Specialized Vulnerability Agent

Flash Cyber is a dedicated cybersecurity model focused on vulnerability discovery and remediation. Integrated into the CodeMender agent, it coordinates multiple Cyber-specialized agents to achieve state-of-the-art results on the CyberGym benchmark at lower cost than larger models. The model addresses the growing gap where AI-driven vulnerability detection outpaces traditional patching pipelines. Access is currently restricted and not available to the general public.

Gemini 4 Pre-training Underway

Google confirmed it has started what it describes as its "most aggressive pre-training run ever," targeting Gemini 4. Based on the company's typical six-month training cadence, a release could arrive by year-end. The article notes that while the three new models deliver immediate cost and efficiency gains — lower token usage, lower pricing, lower per-agent run cost — both the still-unreleased 3.5 Pro flagship and Gemini 4 remain "futures" for now.

Reference: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/?utm_source=tw&utm_medium=social&utm_campaign=og