OpenAI Launches GPT-5 with Reasoning Capabilities, Anthropic Counters with Claude 4 Opus

OpenAI has officially launched GPT-5, marking the most significant model upgrade since GPT-4. The new model introduces native reasoning capabilities that allow it to work through complex problems step by step before answering. Early benchmarks show GPT-5 outperforming GPT-4o on mathematics, coding, and scientific reasoning tasks by significant margins. The model also brings improved multimodal understanding, processing text, images, and audio in a unified architecture.

Anthropic responded within days by announcing Claude 4 Opus, positioned as their most capable model to date. The company claims particular strength in long-context reasoning and tool use, with a 200,000 token context window. Both releases signal that the frontier model race has entered a new phase focused on reasoning depth rather than just scale.

Google DeepMind published research demonstrating AlphaFold 3's ability to model protein interactions with DNA, RNA, and small molecules at atomic accuracy. The system predicts how proteins bind to drug-like molecules, potentially accelerating drug discovery timelines from years to months. DeepMind is making the model available through a free server for non-commercial research, with commercial licensing through Isomorphic Labs.

The European Union AI Act enters its next enforcement phase on August 2nd, 2026. Providers of high-risk AI systems must now comply with requirements for risk management, data governance, transparency, and human oversight. Companies deploying AI in critical infrastructure, education, employment, and law enforcement face the strictest obligations. Penalties reach up to 35 million euros or 7 percent of global turnover.

Microsoft announced a new Copilot+ PC certification program requiring neural processing units capable of 40 trillion operations per second. The specification aims to ensure local AI processing for privacy-sensitive workloads. Major OEMs including Dell, HP, and Lenovo have committed to launching certified devices this quarter. The move reflects a broader industry shift toward on-device AI inference.

Researchers at Stanford released a study showing that large language models can now generate functional code for 85 percent of common programming tasks without human intervention. The study evaluated models on a benchmark of 1,000 real-world GitHub issues across popular repositories. Results suggest AI coding assistants are approaching professional developer productivity for routine tasks.

This article was written by Triweb AI's editorial team based on analysis of today's leading AI news sources.