Google launches Gemini 3.6 Flash and 3.5 Flash-Lite, teases Gemini 4

9to5google.com Jul 21, 2026

Google launches Gemini 3.6 Flash and 3.5 Flash-Lite, teases Gemini 4

Curated by Shubham · TechCapsules · Jul 21, 2026

AI-summarised brief · reviewed before publication

Google has launched Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, alongside a specialized security model, Gemini 3.5 Flash Cyber. Gemini 3.6 Flash improves token efficiency by 17% and lowers pricing to $1.50 per million input tokens. It demonstrates enhanced coding precision, achieving 49% on DeepSWE, and updates its knowledge cutoff to March 2026. Gemini 3.5 Flash-Lite targets high-throughput tasks with a price of $0.30 per million input tokens, significantly outperforming its predecessor in agentic and long-context benchmarks. The new Flash Cyber model, built on the Flash foundation, aims to detect and patch code vulnerabilities at scale. It is currently available to governments and trusted partners via a limited-access pilot program using the CodeMender tool. Both new Flash models are immediately available in the Gemini app, Google Antigravity, AI Studio, and Android Studio. Google confirmed that Gemini 3.5 Pro remains in partner testing. Additionally, the DeepMind team announced it has begun pre-training for the next-generation Gemini 4 model, marking their most ambitious training run to date.

💡 Why It Matters

· The introduction of a dedicated security-focused AI model signals a strategic shift toward automated, large-scale vulnerability management for critical infrastructure.
· By decoupling security tasks from general-purpose models, Google enables specialized, cost-effective defense mechanisms that operate ahead of potential exploits.

AI Capsules Google

Read full article on 9to5google.com

CrunchJ Launches Marketing Reporting Platform with AI Visibility and Conversion Attribution

CrunchJunkie, a Hamburg‑based SaaS firm, unveiled a new marketing reporting platform that merges automated performance dashboards with AI‑visibility tracking. The tool aggregates data from Google Ads, Meta, GA4, Microsoft Ads and other sources into an interface, generates AI‑written executive summaries in a brand’s tone, and schedules white‑label PDF reports, reducing report creation from a full day to roughly ten minutes. It also monitors how eight major AI assistants—including ChatGPT, Gemini, Claude and Perplexity—describe a brand, measuring visibility, sentiment, share of voice and cited sources. By linking this AI visibility to Google Analytics traffic and conversions, the platform quantifies the impact of AI mentions. Plans start at €12 per month and include a GEO audit and a Signals inbox for alerts.

💡 Marketers can finally prove that AI‑driven brand mentions translate into sales, turning speculative visibility into a measurable ROI driver.

nextplatform.com · Jul 21

GenAI Hardware Investments Are Way Ahead Of Model And Platform Revenues

Recent analysis shows that investment in generative AI (GenAI) hardware far outpaces revenue from AI models and platforms. Gartner’s data estimates model spending at $1.4 billion in 2023, rising to $5.7 billion in 2024, but growth slows sharply thereafter, with projected 2026 revenues of $28.3 billion—well below the annualized revenue run rates reported by OpenAI and Anthropic. The disparity highlights a widening gap between capital‑intensive infrastructure spending and the cash flow generated by AI services. Open‑weight models from Chinese firms such as Alibaba Qwen and DeepSeek are gaining traction, potentially diverting customers from higher‑priced closed‑weight offerings. Meanwhile, U.S. and European firms face mounting hardware costs that may outstrip the earnings from their proprietary models, raising concerns about the sustainability of current investment levels.

💡 The hardware‑revenue mismatch threatens to strain cash‑flow for leading AI developers, forcing them to reassess growth strategies amid rising competition from cheaper open‑weight alternatives.

therobotreport.com · Jul 21

Generative Bionics unveils humanoid robot with full-body tactile sensing

Generative Bionics introduced its Gene.01 humanoid at AMD Advancing AI 2026, marking the robot’s U.S. debut. Developed in six months, the platform integrates full‑body tactile skin that senses touch, temperature, proximity and force with a physics‑native AI that treats the robot’s body, mechanics and intelligence as a single system. CEO Daniele Pucci described Gene.01 as a scalable, human‑aware platform that can be customized at the AI, design and end‑effector levels for specific industrial tasks. The first commercial configuration partners with shipbuilder Fincantieri to perform welding in shipyards. Gene.01’s model and digital twin are open‑sourced on PyPI, conda‑forge and ROS, enabling developers to integrate the robot instantly into existing AI and simulation pipelines. Its rapid build underscores fast tactile‑robotic progress today.

💡 By open‑sourcing a touch‑sensitive humanoid, Generative Bionics lowers the barrier for factories to deploy collaborative robots that can learn tasks through force feedback, accelerating adoption of safe human‑robot teamwork.

247wallst.com · Jul 21

China’s New AI Breakthrough Shows Tim Cook’s Risky Final Bet Might Actually…

China’s Cyberspace Administration has officially approved Apple Intelligence for launch in the country, resolving a regulatory standoff that persisted since the feature’s 2024 debut. The clearance relies on strategic partnerships with domestic tech giants Alibaba and Baidu, integrating their respective AI models, Qwen and Baidu’s technology, into Apple’s operating systems. This development validates CEO Tim Cook’s controversial “model-agnostic” strategy, which prioritizes external collaborations over building proprietary frontier models. Following the announcement, Apple shares reached a record high, while Alibaba and Baidu stocks also surged. The approval unlocks access to a critical market where Apple’s sales recently grew 28% year-over-year to $20.5 billion. By leveraging local partners to meet strict Chinese AI regulations, Apple demonstrates a capital-light approach to artificial intelligence. This contrasts sharply with competitors’ heavy infrastructure investments. Although no specific launch date is set, the move highlights Apple’s ability to navigate complex regulatory environments while maintaining its position as a leading smartphone vendor in Greater China.

💡 Apple’s reliance on local partners transforms regulatory compliance from a barrier into a scalable operational advantage. This capital-light model allows the company to deploy AI globally without bearing the massive infrastructure costs that burden its tech rivals.

analyticsinsight.net · Jul 21

10 Technologies Changing Professional Sports

Professional sports are undergoing a significant technological transformation driven by ten key innovations designed to enhance performance, fairness, and fan engagement. Wearable trackers and GPS monitoring systems allow athletes to optimize training loads and reduce injury risks through precise data on heart rate, speed, and movement. Artificial intelligence analytics process vast datasets to identify tactical improvements and strategic opportunities for coaching staff. Officiating accuracy has improved with Video Assistant Referee systems and Hawk-Eye tracking, which utilize multiple camera angles to review controversial decisions and determine ball trajectories. Smart equipment sensors embedded in gear collect real-time performance metrics during competitions. Meanwhile, virtual reality training provides immersive environments for safe skill development, and biomechanics motion capture helps refine techniques. Finally, fan engagement platforms leverage augmented reality and interactive apps to create immersive experiences, strengthening participation both in stadiums and through digital broadcasts, fundamentally changing how sports are played, officiated, and consumed globally.

💡 These tools shift sports from intuition-based decisions to data-driven precision, fundamentally altering competitive advantages. By quantifying every aspect of play, teams and leagues are redefining the boundaries of human performance and viewer immersion.

infoq.com · Jul 21

Presentation: Engineering AI for Creativity and Curiosity on Mobile

Bhavuk Jain, a Tech Lead at Google, presented on engineering artificial intelligence for creativity and curiosity on mobile devices at QCon San Francisco. He detailed the technical challenges of translating foundational AI models into scalable mobile products, specifically citing AI Wallpapers and Circle to Search. Jain explained the implementation of robust runtime guardrails, model fine-tuning, and seamless operating system integration. He addressed engineering leaders on balancing user experience constraints with model latency and infrastructure costs to ensure safe, reliable AI deployment. His work, which has been adopted by leading global device manufacturers and integrated into the Android Open Source Project, demonstrates how applied AI initiatives can enhance Pixel devices. The presentation provided practical insights for technical team leads, architects, and project managers on delivering high-quality AI features while managing complex engineering trade-offs in mobile environments.

💡 Integrating AI directly into the operating system shifts the competitive advantage from standalone apps to foundational platform capabilities. This architectural shift forces developers to prioritize efficiency and safety over raw model power.