Google Releases Three Gemini Models & Starts Gemini 4 Pre-training

Three New Gemini Models Released Simultaneously

Google DeepMind announced three new Gemini models in a single release, each targeting different production needs for AI agents: a stronger flagship, a faster lightweight model, and a security-focused specialist.

Gemini 3.6 Flash: Fewer Tokens, Lower Cost, Higher Benchmarks

Gemini 3.6 Flash is positioned as the new workhorse. According to Artificial Analysis Index, it uses 17% fewer output tokens than 3.5 Flash overall, and up to 65% fewer on the DeepSWE coding benchmark. It also requires fewer reasoning steps and tool calls, making it both faster and cheaper.

Pricing drops to $1.50 per million input tokens and $7.50 per million output tokens, undercutting 3.5 Flash. Despite the lower cost, benchmark scores improve significantly:

DeepSWE (coding): 37% → 49%

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

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

GDPVal-AA v2 (knowledge work): 1421 tasks completed, 70+ more than predecessor

Demos show 3.6 Flash orchestrating multi-agent code migrations, generating 3D workflows via Gemini Canvas, and acting as an AI interaction designer for immersive theme studios.

Gemini 3.5 Flash-Lite: Speed and Cost for High-Volume Workloads

Flash-Lite targets high-throughput scenarios like bulk document processing and agent search. It outputs at 350 tokens/second — the fastest in the 3.5 series — at $0.30 per million input tokens and $2.50 per million output tokens.

Surprisingly, this lightweight model outperforms the larger Gemini 3 Flash on key benchmarks:

SWE-Bench Pro: 54.2% vs 49.6%

OSWorld-Verified: 74.0% vs 65.1%

The recommended architecture pairs 3.6 Flash as the "brain" for task decomposition with multiple Flash-Lite instances as "workers" for parallel execution. A demo generated 25 production-ready web designs in seconds. Flash-Lite is already live in the Gemini App and rolling into Google Search.

Gemini 3.5 Flash Cyber: Specialized Vulnerability Detection and Remediation

Flash Cyber is a dedicated cybersecurity model integrated into the CodeMender agent. Multiple Cyber agents collaborate to find and fix vulnerabilities, achieving state-of-the-art on the CyberGym benchmark at lower cost than larger models. The model addresses the growing gap where AI discovers vulnerabilities faster than existing systems can patch them. Access is currently restricted and not available to the general public.

Gemini 4 Pre-training Underway: "Most Aggressive Ever"

On the same day, Google confirmed it has begun pre-training for Gemini 4, calling it the most aggressive pre-training run in DeepMind history. Based on Google's typical six-month training cycle, Gemini 4 could appear by year-end. The article notes that while flagship models and Gemini 4 remain "futures," the three released models deliver immediate practical value: lower prices, reduced token consumption, and lower per-agent operating costs.

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