US EditionPublished • loading... • Updated
New ChatGPT 5.4 Model Is 'Built for Agents.' Will It Lure Back Claude Converts?
- OpenAI released GPT-5.4 on Thursday as a new frontier model for professional work, rolling it out in ChatGPT and Codex for Plus, Team, and Pro subscribers and via the API.
- Designed for complex workflows, GPT-5.4 supports a 1 million-token context window, built-in computer-use capabilities, and six key improvements in reasoning, tool use, and long-running task execution, OpenAI says.
- Benchmarking shows GPT-5.4 scored 83% on the GDPval test, is 33% less likely to make false claims, and leads Mercor APEX-Agents and OSWorld/WebArena benchmarks with improved spreadsheet modeling and visual understanding.
- The launch gives enterprise customers GPT-5.4's autonomous agentic features and ChatGPT for Excel and Google Sheets beta, alongside new FactSet, MSCI, Third Bridge, and Moody's integrations.
- Meanwhile, the company keeps GPT-5.3 Instant as the default while planning to offer GPT-5.4 via API on Friday for Plus, Team, and Pro subscribers.
Insights by Ground AI
22 Articles
22 Articles
Coverage Details
Total News Sources22
Leaning Left2Leaning Right1Center7Last UpdatedBias Distribution70% Center
Bias Distribution
- 70% of the sources are Center
70% Center
L 20%
C 70%
Factuality
To view factuality data please Upgrade to Premium
