Triweb AI
Back to AI Insights
Latest25 Aug 2026Triweb AI3 min read

OpenAI Announces GPT-5 Preview Access for Enterprise Customers

OpenAI rolls out GPT-5 preview to enterprise partners while Anthropic updates Claude 3.5 Sonnet and Google DeepMind publishes research on adaptive compute scaling for smaller models.

OpenAI Announces GPT-5 Preview Access for Enterprise Customers

OpenAI has begun rolling out preview access to GPT-5 for select enterprise customers, marking the first major model release since GPT-4o in May 2024. The company confirmed the preview program in a brief blog post, noting that early partners include firms in financial services, legal technology, and software development. OpenAI says the new model improves on complex reasoning, multi-step coding tasks, and long-context accuracy, areas where GPT-4o still struggles in production workloads.

Anthropic quietly released an updated version of Claude 3.5 Sonnet over the weekend, improving its performance on coding benchmarks and reducing hallucination rates on technical documentation tasks. The update was not accompanied by a formal announcement but appeared in the model selection dropdown for API users. Developers testing the new version report noticeably better results on repository-wide refactoring and legacy code migration tasks.

Google DeepMind published a research paper detailing a new approach to test-time compute scaling that allows smaller models to match larger ones on specific reasoning tasks by allocating more inference compute per query. The technique, called adaptive compute allocation, dynamically adjusts the number of reasoning steps based on problem difficulty. The paper shows a 7B parameter model achieving performance comparable to a 70B model on math and logic benchmarks when given 10x inference budget.

Microsoft announced expanded Copilot integration across the Power Platform, adding autonomous agents that can execute multi-step workflows across Dynamics 365, Dataverse, and custom connectors. The agents operate within defined guardrails and require human approval for destructive actions. Microsoft positions this as a step toward agentic enterprise software where routine operations are delegated to AI while humans handle exceptions and strategy.

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

Share this article

Get AI insights delivered weekly

Like what you read? Join the newsletter and receive the latest AI news, analysis, and practical insights straight to your inbox every week - written by Triweb AI.

Free AI Implementation Starter Kit