What the monitor gets to read and the capability of the monitor matters. Stronger monitors that can read CoTs and use more test-time compute get much better fast. Also, post-hoc follow-ups (by asking the model to elaborate) often surface previously unspoken thoughts and boost
@openai
-
Chain-of-thought monitoring as complementary mechanistic interpretability approach
By
–
We view chain-of-thought monitoring as complementary to mechanistic interpretability, not as a replacement for it. Because we believe that chain-of-thought monitoring is incredibly useful as a window into a model’s brain and could be a loadbearing layer in a scalable control
-
Chain-of-Thought Monitoring for Better Model Oversight
By
–
Monitoring a model’s chain-of-thought is far more effective than watching only its actions or final answers. The more a model “thinks” (longer CoTs), the easier it is to spot issues.
-
RL Reasoning Tradeoff: Monitorability vs Inference Compute
By
–
RL at today’s frontier doesn’t seem to wreck monitorability and can help early reasoning steps. But there’s a tradeoff: smaller models run with higher reasoning effort can be easier to monitor at similar capability — at the cost of extra inference compute (a “monitorability
-
Measuring Chain-of-Thought Monitorability in AI Models
By
–
To preserve chain-of-thought (CoT) monitorability, we must be able to measure it. We built a framework + evaluation suite to measure CoT monitorability — 13 evaluations across 24 environments — so that we can actually tell when models verbalize targeted aspects of their
-
GPT-5.2-Codex Advances Agentic Coding and Cybersecurity Standards
By
–
GPT-5.2-Codex is now available in Codex. It sets a new standard for agentic coding in real-world software development and defensive cybersecurity. It also delivers more reliable performance on complex tasks and scales effectively across large projects.
-
New US AI Data Center in Wisconsin Creates 5000+ Jobs
By
–
That’s why we’re building new AI infrastructure in the U.S., including a data center in Wisconsin that will create 4000+ skilled construction jobs, 1000+ long-term jobs, and be energy and water positive for the community (contributing more than it takes) — all while expanding the
-
Compute Power: Foundation for AI Healthcare and Scientific Breakthroughs
By
–
Beyond image generation, we know people want more from AI like healthcare improvements and scientific discovery. Compute is how we get it. If America doesn't build it, others will.
-
Image Generation Launch Drives 32% WAU Growth, Compute Needs Surge
By
–
Compute enabled our first image generation launch (and a +32% jump in WAU over the following weeks) as well as our latest image generation launch yesterday. We have a lot more coming… and need a lot more compute. pic.twitter.com/rHfQv1aLKS
— OpenAI (@OpenAI) 17 décembre 2025Compute enabled our first image generation launch (and a +32% jump in WAU over the following weeks) as well as our latest image generation launch yesterday. We have a lot more coming… and need a lot more compute.
-
ChatGPT Introduces New Images Surface Feature
By
–
We’re also introducing a new Images surface within ChatGPT. Just tap “Images” in the sidebar and have fun. And make sure you update your app!