🚀 Google processes nearly one quadrillion tokens in June 🚀
Exponential growth of reasoning models usage confirmed

Google's AI systems processed over 980 trillion tokens in June—more than double the amount in May, according to Google product manager Logan Kilpatrick and DeepMind CEO Demis Hassabis. The massive scale achievement reflects exponential growth of AI usage across Google's platforms.

Demis Hassabis announced the milestone on X: "You know what's cool... a quadrillion tokens. We processed almost 1,000,000,000,000,000 tokens last month, more than double the amount from May." This represents one of the largest AI processing achievements publicly disclosed.

🧠 Alibaba's Qwen3 thinking model tops OpenAI and Gemini 🧠
New Open-source reasoning AI with enterprise-friendly licensing

Alibaba's Qwen Team has released Qwen3-235B-A22B-Thinking-2507, a new open-source reasoning model that leads or closely trails top-performing models across several major benchmarks, outperforming even leading proprietary options from OpenAI and Google.

Released under Apache 2.0 license, the model offers enterprises full flexibility for commercial deployment, modification, and self-hosting without restrictions. Available for free download on Hugging Face and ModelScope, with API pricing at $0.70 per million input tokens and $8.40 per million output tokens.

🔬 ASI-ARCH: First AI system to autonomously discover neural architectures 🔬
Shanghai researchers establish first empirical scaling law for scientific discovery

Researchers from Shanghai Jiao Tong University have unveiled ASI-ARCH, the first demonstration of Artificial Superintelligence for AI research that autonomously conducts end-to-end scientific research in neural architecture discovery. The system represents a paradigm shift from automated optimization to automated innovation.

ASI-ARCH conducted 1,773 autonomous experiments over 20,000 GPU hours, culminating in the discovery of 106 innovative, state-of-the-art linear attention architectures. Like AlphaGo's Move 37, these AI-discovered architectures demonstrate emergent design principles that systematically surpass human-designed baselines.

INTERESTING TO KNOW

⚡ Singapore startup's brain-inspired AI delivers 100x faster reasoning ⚡

Sapient Intelligence, a Singapore-based research lab, has unveiled the Hierarchical Reasoning Model (HRM), a brain-inspired architecture that achieves deep reasoning with just 1,000 training examples and delivers up to 100x faster task completion than traditional LLMs.

🎨 Google launches Opal for no-code AI apps 🎨

Opal enables users to create workflows by chaining together prompts, AI model calls, and tools using conversational natural language commands. The platform translates instructions into visual workflows, giving users fine-grained control without ever needing to see code.