Key Highlights
- Google unveiled a trio of Gemini AI models: 3.6 Flash, 3.5 Flash-Lite, and the security-oriented 3.5 Flash Cyber
- The Gemini 3.6 Flash model delivers 17% reduction in output token consumption compared to earlier versions, with pricing set at $1.50 per million input tokens
- Gemini 3.5 Flash Cyber targets cybersecurity applications and is exclusively accessible to governmental entities and vetted partners through CodeMender
- The lightweight 3.5 Flash-Lite variant achieves 350 tokens per second output speed at $0.30 per million input tokens
- These model releases precede Alphabet’s earnings announcement, with GOOGL shares declining 0.66% to reach $349.66
On Tuesday, Alphabet unveiled three cutting-edge Gemini AI models, timing the announcement strategically ahead of its upcoming quarterly financial results. Trading activity showed GOOGL shares at $349.66, reflecting a 0.66% decrease.
The trio of releases — designated as Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — specifically targets developers and enterprise clients seeking to deploy artificial intelligence solutions at substantial scale.
Taking center stage is Gemini 3.6 Flash. According to Google’s performance data, this model surpasses 3.5 Flash across coding tasks, knowledge-intensive operations, and multimodal applications, all while consuming 17% fewer output tokens. The pricing structure stands at $1.50 per million input tokens and $7.50 per million output tokens.
Today we’re expanding the Gemini family with three new models built to be faster, more token efficient, and reliable at scale.
Meet the new Gemini models ↓ pic.twitter.com/hVYNQqxgjd
— Google (@Google) July 21, 2026
Benchmark testing reveals impressive gains: 3.6 Flash achieves a 49% score on DeepSWE, marking significant improvement from the 37% registered by 3.5 Flash. MLE Bench performance climbed to 63.9% from the previous 49.7%. Computer utilization capabilities on OSWorld-Verified also advanced, reaching 83.0% compared to 78.4%.
Early adopters include prominent names such as Figma, Harvey, Hebbia, and JetBrains. Both Hebbia and Harvey have highlighted the model’s exceptional multimodal capabilities, especially regarding document interpretation and visual data analysis.
Gemini 3.5 Flash-Lite: Optimized for Velocity and Affordability
Designed for high-volume operations demanding rapid processing, the 3.5 Flash-Lite model delivers 350 output tokens per second. Its competitive pricing sits at $0.30 per million input tokens and $2.50 per million output tokens.
Flash-Lite demonstrates substantial performance advantages over the previous 3.1 Flash-Lite generation, particularly in agentic workflows and programming tasks. Terminal-Bench 2.1 testing shows a 54% score against 31%. SWE-Bench Pro results indicate 54.2%, outpacing the earlier 3 Flash model’s 49.6%.
Among initial enterprise users, Palo Alto Networks and Ramp have adopted the model, emphasizing its velocity and economic advantages for expanding operational workflows.
Security-Specialized Model Introduced
Gemini 3.5 Flash Cyber represents a specialized security solution, refined from the 3.5 Flash foundation to identify and remediate software vulnerabilities. It functions within Google’s CodeMender agent ecosystem, where multiple Flash Cyber agents collaborate to generate comprehensive security assessments.
Google has adopted a measured rollout strategy. Access to this model remains restricted to governmental agencies and authorized partners through a controlled pilot program, acknowledging the potential dual-use implications of the technology.
Performance metrics on the CyberGym benchmark indicate Flash Cyber delivers what Google characterizes as frontier-level competitive capability.
This cybersecurity initiative positions Google in closer competition with OpenAI and Anthropic, both having recently introduced solutions targeting security professionals and vulnerability identification.
Google’s premier Gemini 3.5 Pro model remains under partner evaluation. Initially anticipated for June release, no definitive launch timeline has been announced. The company has also disclosed that pre-training for Gemini 4 has commenced.
Both 3.6 Flash and 3.5 Flash-Lite are immediately accessible through Google AI Studio, Android Studio, Gemini Enterprise, and the Gemini application. Flash-Lite is additionally being integrated into Google Search.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
