Google has introduced a new AI security model called Gemini 3.5 Flash Cyber, designed as a cost-efficient alternative to larger AI systems. This model is built upon Gemini 3.5 Flash and will be available first to governments and trusted partners via CodeMender, Google's security-focused coding agent.
Google has also released three new Gemini models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. These models are designed for speed, cost, and high-volume agentic work. Gemini 3.6 Flash is the new default workhorse, targeting coding, knowledge work, and multimodal tasks.
Meanwhile, executives from OpenAI and Meta seem to be mocking Google's recent release of Gemini 3.6 Flash. Meta's AI Chief, Alexandr Wang, posted 'gemini who?' on social media, while OpenAI's Tibo Sottiaux responded with 'Hope it finishes one day too!' The exchange suggests that Google's competitors are paying close attention to its AI developments.
Other AI-related news includes BlueQubit receiving $1.5 million in DOE Genesis Mission grants to advance AI-driven quantum error correction. Additionally, Poolside has released Laguna S 2.1, an open-weight agentic coding model that offers high performance at a lower cost.
Neura Robotics is partnering with RWTH Aachen University to establish a physical AI training facility in Germany. Capital One has open-sourced Vulnhunter, an agentic AI security tool designed to combat cyber threats.
Key Takeaways
['Google launches Gemini 3.5 Flash Cyber, a cost-efficient AI security model.', 'Google releases new Gemini models for speed, cost, and high-volume agentic work.', "OpenAI and Meta executives appear to mock Google's Gemini release.", 'BlueQubit receives $1.5 million for AI-driven quantum error correction.', 'Poolside releases Laguna S 2.1, an open-weight agentic coding model.', 'Neura Robotics partners with RWTH Aachen University for AI training facility.', 'Capital One open-sources Vulnhunter AI security tool.', "NVIDIA's stock holds at $207.92 as AI supply concerns weigh on the market.", 'Fei-Fei Li discusses spatial AI and its applications.', 'MUX enables high-bandwidth and compact reasoning.']Google Launches Cheaper AI Security Model
Google has introduced a new AI security model called Gemini 3.5 Flash Cyber, which is designed to be a cost-efficient alternative to larger AI systems. This model is built upon Gemini 3.5 Flash and will be available first to governments and trusted partners via CodeMender, Google's security-focused coding agent. The new model can scan more code paths and find vulnerabilities quickly and at a lower cost.
Google Unveils New Gemini Models for Speed and Efficiency
Google has released three new Gemini models: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. These models are designed for speed, cost, and high-volume agentic work. Gemini 3.6 Flash is the new default workhorse, targeting coding, knowledge work, and multimodal tasks. The models offer improved efficiency and lower pricing, making them suitable for various applications.
OpenAI and Meta Execs Appear to Mock Google's Gemini Release
Executives from OpenAI and Meta seem to be mocking Google's recent release of Gemini 3.6 Flash. Meta's AI Chief, Alexandr Wang, posted 'gemini who?' on social media, while OpenAI's Tibo Sottiaux responded with 'Hope it finishes one day too!' The exchange suggests that Google's competitors are paying close attention to its AI developments.
MUX Enables High-Bandwidth and Compact Reasoning
Researchers have proposed MUX, a method for high-bandwidth and compact reasoning based on distillation of discrete reasoning into continuous multiplexed tokens. MUX allows for lossless superposition and prevents shortcut behaviors caused by latent collapse. The approach shows promising results across various evaluation settings.
Alphabet Earnings D-Day with NVDA at $207
Alphabet is set to report its Q2 earnings, and its stock price is closely watched. Meanwhile, NVIDIA's stock holds at $207.92 as AI supply concerns weigh on the market.
BlueQubit Receives $1.5M for AI-Driven Quantum Error Correction
BlueQubit and its partners have been awarded $1.5 million in DOE Genesis Mission grants to advance AI-driven quantum error correction. The funding will support the development of fault-tolerant quantum computing systems.
Course Helps Publishers Navigate AI Responsibly
A new online course and live seminar series will provide practical information on AI usage in the global publishing industry. The program aims to help publishers understand, accept, and implement AI in their workflow responsibly.
Fei-Fei Li Discusses Spatial AI and Its Applications
Fei-Fei Li, co-founder of World Labs, discussed spatial AI and its potential applications. World Labs has shipped its first commercial product, Marble, which generates explorable 3D environments. The company has raised $1.23 billion in funding.
Poolside Releases Laguna S 2.1 for Coding Efficiency
Poolside has released Laguna S 2.1, an open-weight agentic coding model that offers high performance at a lower cost. The model scores well on various benchmarks, including SWE-Bench Multilingual and Terminal-Bench 2.1. Laguna S 2.1 uses a sparse activation approach, making it efficient for deployment.
Neura Robotics Expands Training Network with German Neura Gym
Neura Robotics is partnering with RWTH Aachen University to establish a physical AI training facility in Germany. The site will enable robots to learn from real-world interactions and human collaboration. Neura plans to develop multiple training facilities globally.
Capital One Open-Sources Vulnhunter AI Security Tool
Capital One has open-sourced Vulnhunter, an agentic AI security tool designed to combat cyber threats. The tool uses machine learning to identify and classify potential threats in real-time. By making Vulnhunter available on GitHub, Capital One aims to promote collaboration and improve cybersecurity.
Sources
- Google launches a cheaper alternative to large AI security models like Mythos
- Google Releases Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber: A Cheaper, More Token-Efficient Flash Tier Built for Agentic Workloads
- OpenAI, Meta Execs Appear To Mock Google After Release Of Gemini Flash 3.6
- MUX: Continuous Reasoning via Multiplexed Tokens
- Trader Claude's: July 22, 2026, Alphabet Earnings D-Day, NVDA at $207
- BlueQubit and Partners Awarded $1.5M in DOE Genesis Mission Grants to Advance AI-Driven Quantum Error Correction
- Announcing a Course to Help Publishers Navigate the AI Age Responsibly
- Fei-Fei Li on Spatial AI: Marble, $1.23B, and the Language Gap
- Poolside Releases Laguna S 2.1, an Open-Weight Agentic Coding Model Punching Above Its Weight Class on SWE-Bench Multilingual
- Neura Robotics Expands its Global Training Network for Physical AI with New German Neura Gym
- FF News
Comments
Please log in to post a comment.