Emergent introspective awareness in large language models
Interpretability Signs of introspection in large language models Oct 29, 2025 Read the paper Have you ever asked an AI model what’s on its mind? Or to explain how it came…
Search
Making Claude a chemist
…We measured three Claude models (Opus 4.7, Opus 4.6, Sonnet 4.6) against ChemDraw and MestReNova on 20 compounds drawn from synthetic chemistry preprints published after the models’ training cutoff…
The assistant axis: situating and stabilizing the character of large language models
…We tracked how model activations moved along the Assistant Axis throughout each conversation. The pattern was consistent across the models we tested. While coding conversations kept models firmly in Assistant territory throughout…
Paving the way for agents in biology
…Run-to-run variability was largely eliminated, and the performance gap between models narrowed dramatically. In other words, adding a deterministic retrieval layer made model choice much less important . This is especially…
Mapping AI-enabled cyber threats: Insights from the LLM ATT&CK Navigator
…Instead, the differentiator will become the scaffolding—the surrounding code, architecture, and tooling that makes AI models more capable—that actors build around the model so they can chain together attack stages…
Eval awareness in Claude Opus 4.6’s BrowseComp performance
…We have updated the model cards for both Claude Opus 4.6 and Claude Sonnet 4.6. For the Opus 4.6 multi-agent configuration described in this report, the run we…
AI agents find smart contract exploits
…We hope that this post helps to update defenders' mental model of the risks to match reality—now is the time to adopt AI for defense. If you want to contribute to…
Values in the wild: Discovering and analyzing values in real-world language model interactions
…An initial update An early update on what we've learned from Project Glasswing.
Focus areas for The Anthropic Institute
…Markets steer the direction of model improvement according to private return, but can we improve how models perform to address social externalities? Related content Teaching Claude why New research on how we…
Estimating AI productivity gains
…minor documentation updates and one is a critical infrastructure change, simply counting the number of these tasks performed with Claude misses the point. Not only that, but as model capabilities improve, we…