Google Releases Gemini 3.6 Flash Trio While Competitors Launch New Models

모델 출시 | Wed Jul 22 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 5 sources

Google unveiled three Gemini Flash models, Kimi K3 open model emerged, and Qwen-Image-3.0 was announced.

Analysis

[Google DeepMind] released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber simultaneously ^[2]

• Workhorse model 3.6 Flash
• Most affordable 3.5 Flash-Lite
• Security-specialized 3.5 Flash Cyber
• Gemini 3.5 Pro not included

[Gemini 3.6 Flash] announced workhorse model with improved coding and multimodal performance ^[3]

• 17% reduction in output tokens
• Input $1.50/1M
• output $7.50/1M
• DeepSWE 49% vs 37%
• MLE Bench 63.9% vs 49.7%

[Gemini 3.5 Flash Cyber] released cybersecurity-specialized model ^[1]

• Integrated into CodeMender agent
• Limited access for government and trusted partners
• Cost-effective alternative to Anthropic Mythos
• Found 55 unique issues in V8 engine

[Gemini 3.5 Flash-Lite] launched most cost-effective ultra-lightweight model ^[3]

• 350 output tokens per second
• Best cost-efficiency in 3.5 class
• Enhanced agentic workflow performance

[Kimi K3] released frontier-class open model rivaling Fable 5 ^[4]

• SWE benchmark 92.4% (Fable 92.6%)
• Up to 50x cost efficiency vs Fable
• 93% accuracy when routing
• Validated on 1
• 030 agentic tasks

[Qwen] announced Qwen-Image-3.0 image generation model ^[5]

• Emphasis on rich content
• Authentic detail expression
• Image generation based on deep knowledge

Sources

• [1] Google launches a cheaper alternative to large AI security models like Mythos - The Verge AI
• [2] Google releases three new Gemini models — but no 3.5 Pro - TechCrunch AI
• [3] Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber - Hacker News
• [4] Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA - Hacker News
• [5] Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge - Hacker News