OpenAI says GPT-5.4 reduces hallucinations and factual errors, reporting a 33% drop in individual claim errors versus GPT-5.2 and an 18% reduction in overall errors.

Sam Altman acknowledged that the contract announcement may have seemed rushed, as critics from users and staff persisted and some departures were reported over the issue.

The piece includes a brief author bio and a note on confirming a public display name before commenting, signaling publication context and engagement requirements.

The article flags breaking news status and promises to provide updates as more information becomes available.

Agentic adoption is moving quickly, pressuring competitors and driving enterprise adoption through integrations with spreadsheets and data partners; broader rollout, including full iOS access, remains to be seen.

Ethical backlash over the Pentagon contract poses a challenge not captured by benchmarks, highlighting a gap between safety commitments and defense terms as questioned by OpenAI leadership.

There is ongoing uncertainty about whether AI will augment or displace professionals, urging readers to consider adaptation strategies as capabilities advance.

Customer impact centers on financial institutions using GPT-5.4 to improve due diligence, underwriting, and client engagement, with ongoing learning as deployments scale.

The Information cites a knowledge-holder and notes the report was AI-assisted and editor-reviewed.

The piece frames a rapid OpenAI-Anthropic competition with aggressive benchmarks and feature launches amid broader debates on autonomous systems and deployment.

OpenAI has released GPT-5.4, a next-generation model designed for professional use with enhanced reasoning, coding, and agentic abilities, available to ChatGPT Plus, Team, Pro subscribers, and via API, including a Thinking variant for mid-task adjustments and a Pro enterprise version.

GPT-5.4 shows strong performance gains, including 75% on the OSWorld-Verified benchmark (above the human baseline) and a 1-million-token context window to support extended workflows, while remaining more token-efficient.