OpenAI Releases GPT-6.1 Terra With Improved Safety but Persistent Exploit Vulnerabilities

OpenAI unveiled GPT-6.1 Terra on October 26, 2026, a faster and significantly cheaper model designed for high-volume automation tasks. While the model demonstrates a 42% reduction in severity-level safety flags compared to GPT-6 Sol and 64% fewer than GPT-5.6 Sol, independent safety evaluations reveal concerning gaps in authorization and restriction enforcement that warrant careful deployment consideration.