started · updated
Google, OpenAI, and Anthropic announce major AI model updates
Several major artificial intelligence developers have released significant updates to their model capabilities and security features.
Google has rolled out video generation capabilities for Gemini Omni 1.2 Flash, supporting clips up to 15 seconds and real-time analysis of videos up to two hours long. The company also reduced 720p API pricing by 30 percent and introduced a new Interactions API for in-dialogue edits.
OpenAI launched GPT-6 Cyber, a large language model optimized for cybersecurity tasks such as vulnerability diagnosis, penetration testing, and malware analysis. During early testing, the model identified 948 vulnerabilities in the Linux kernel. Access is currently limited to approved users through a qualification process to prevent misuse.
Anthropic introduced FlashProve, a technology that uses zero-knowledge proofs (zkSNARK) to cryptographically verify an AI model's reasoning process. This allows third parties to mathematically confirm which model performed a specific inference, aiding compliance with regulations like the EU AI Act. Additionally, Anthropic added a Message Threads feature to the Claude API, allowing the server to automatically maintain conversation context for developers.
Entities
Anthropic · GPT-6 Cyber · Gemini Omni 1.2 Flash · Google · OpenAI