Crimson AI NewsA CrimsonLingua Network service
EN ع
← Back to news
Google DeepMind

Google DeepMind Unveils Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

AI By Crimson AI Google DeepMind 21 July 2026 · 15:16 30 views
Share: X Telegram

Google DeepMind introduces three new Gemini models aimed at improving efficiency, latency, and reliability for AI agents, with 3.6 Flash offering 17% fewer output tokens and lower cost than its predecessor.

Google DeepMind Unveils Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Key points

Google DeepMind has announced the release of three new models in its Gemini Flash series: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. These models are designed to meet the growing demands of developers building production AI agents, offering higher token efficiency, lower latency, and more reliable performance.

Gemini 3.6 Flash is positioned as the workhorse model, delivering improved coding, knowledge work, and multimodal capabilities. According to the Artificial Analysis Index, it reduces output token usage by 17% compared to 3.5 Flash, and on benchmarks like DeepSWE, it shows up to 65% reduction in token usage. It is priced at $1.50 per million input tokens and $7.50 per million output tokens, making it more cost-effective than its predecessor. Performance gains include 49% vs. 37% on DeepSWE, 63.9% vs. 49.7% on MLE Bench, and 83.0% vs. 78.4% on OSWorld-Verified. The model also features enhanced Frontier Safety safeguards against CBRN and cyber offense misuses.

Gemini 3.5 Flash-Lite is the fastest model in the 3.5 series, delivering 350 output tokens per second according to Artificial Analysis. It is priced at $0.30 per million input tokens and $2.50 per million output tokens. It significantly outperforms 3.1 Flash-Lite on agentic tasks, with scores of 54% vs. 31% on Terminal-Bench 2.1, 72.2% vs. 60.1% on GDM-MRCR v2, and 1140 vs. 642 on GDPval-AA v2. It even surpasses 3 Flash on several benchmarks, including SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%).

Gemini 3.5 Flash Cyber is a specialized model fine-tuned for cybersecurity vulnerability detection and patching. It operates within the CodeMender agent system, using multiple agents to produce combined reports. It achieves competitive performance on the CyberGym benchmark. Due to dual-use concerns, it will be available exclusively to governments and trusted partners through a limited-access pilot program.

Both 3.6 Flash and 3.5 Flash-Lite are available starting today via Google AI Studio, Android Studio, Gemini Enterprise Agent Platform, and the Gemini app. 3.5 Flash-Lite is also rolling out in Google Search. Google DeepMind also noted that Gemini 3.5 Pro is currently being tested with partners and that pre-training for Gemini 4 has begun.

ModelPrice (Input/1M tokens)Price (Output/1M tokens)Output Speed (tokens/s)Key Benchmark Improvements
3.6 Flash$1.50$7.50N/ADeepSWE 49% (vs 37%), MLE Bench 63.9% (vs 49.7%), OSWorld 83.0% (vs 78.4%)
3.5 Flash-Lite$0.30$2.50350Terminal-Bench 54% (vs 31%), GDM-MRCR 72.2% (vs 60.1%), GDPval-AA 1140 (vs 642)
3.5 Flash CyberN/A (limited access)N/AN/ACompetitive on CyberGym
Source
Google DeepMind · Google DeepMind
Related news
Google DeepMind
Google DeepMind 21 Aug 2026

From Atari to EVE Online: Google DeepMind Marks 15 Years of AI Research in Games

Google DeepMind reflects on 15 years of AI research in games, from Atari to EVE Online, highlighting milestones like AlphaGo and S...

10
Google DeepMind
Google DeepMind 13 Aug 2026

Google DeepMind Unveils Gemini 3.7 Flash: Smarter, Faster, and Half the Price

Gemini 3.7 Flash delivers major gains in coding, web development, and knowledge work, with an introductory price half that of its...

30
Google DeepMind
Google DeepMind 12 Aug 2026

Google DeepMind Launches Sign Language AI in Consumer Products

Google DeepMind introduces SL2T, a multilingual sign-language-to-text model powering sign-to-text dictation in Gboard and Live Tra...

23