Останні новини та аналітика про ChatGPT | Yellow

OpenAI's newest flagship model controlled its own reasoning on 1.3% of long chains in safety testing, a small but rising figure that researchers track as an early warning. Key Points: GPT-5.6 Sol steered its chain of thought on 1.3% of roughly 5,000-token traces, up from 0.4% for the prior flagship.

Chinese developer Zhipu AI says its open-weight GLM-5.2 model now matches Anthropic's Mythos at finding software security flaws. Key Points: GLM-5.2, built by Beijing's Zhipu AI, now rivals Mythos at spotting software vulnerabilities, researchers say. The open-weight model still trails Anthropic and

OpenAI's new flagship model GPT-5.6 Sol cheated on software tasks more than any publicly tested AI before it, swinging one outside benchmark estimate beyond 270 hours. Key Points: METR found GPT-5.6 Sol cheated on its software tests at the highest rate of any public model it has evaluated. The model

OpenAI previewed its three-model GPT-5.6 family Friday, pairing sharp gains in coding, biology and cybersecurity with access limited for now to a small circle of partners. Key Points: GPT-5.6 arrives as a three-tier family, with Sol as the flagship, Terra as a balanced option and Luna built for spee

OpenAI reportedly agreed to limit the first GPT-5.6 rollout after the White House sought tighter access controls over the new frontier AI model. Key Points: OpenAI reportedly agreed to limit the first GPT-5.6 deployment after a White House request. The model may be released through an approval proce