GPT-5.1 Instant and GPT-5.1 Thinking System Card Addendum (openai.com)

OpenAI announced two updates to its GPT‑5 family: GPT‑5.1 Instant (gpt-5.1-instant) and GPT‑5.1 Thinking (gpt-5.1-thinking). GPT‑5.1 Instant is tuned to be more conversational with improved instruction following and an "adaptive reasoning" capability that lets the model decide when to engage additional internal reasoning before replying. GPT‑5.1 Thinking takes that further by dynamically adapting how much “thinking time” it uses per question, trading off latency and compute for higher-quality responses when needed. Users typically won’t need to pick models directly because GPT‑5.1 Auto will route queries to the best-suited variant. The addendum updates baseline safety metrics while keeping most mitigations from the GPT‑5 system card. Crucially, pre-deployment evaluations have been expanded to include mental-health scenarios (e.g., signs of isolated delusions, psychosis, or mania) and “emotional reliance” risks (outputs that could foster unhealthy attachment to the assistant). For the AI/ML community this signals continued emphasis on dynamic compute strategies for adaptive reasoning and stronger, more specific safety auditing for sensitive conversational contexts — both important for deploying capable models responsibly and for research on latency/accuracy trade-offs and human-centered risk mitigation.