GPT-6.1 Sol cuts factual-error responses by about 32% versus GPT-6 Sol
On deliberately difficult factuality prompts, OpenAI Developers says GPT-6.1 Sol reduces responses containing factual errors by about 32% compared with GPT-6 Sol when both run at low reasoning effort.
In OpenAI Developers’ automated safety review evaluation, the team observed no attempts to bypass the safety reviewer.