Project Fetch: Phase two
…perform sophisticated (and amusing) tasks with an off-the-shelf robotic quadruped (henceforth, a robodog). We called this Project Fetch. We found that access to our state-of-the-art model at…
Search
Measuring LLMs' impact on N-day exploits
…On Patch Tuesday, the patched binaries are posted to the Microsoft Update Catalog and a short advisory for each bug appears in the Security Update Guide . Setup We evaluated our models on…
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…
Making Claude a chemist
…The model performance should thus be read as indicative rather than precise. Second, on the densest inverse targets, without the starting material as an additional input, the model could loop through its…
Eval awareness in Claude Opus 4.6’s BrowseComp performance
Engineering at Anthropic Eval awareness in Claude Opus 4.6’s BrowseComp performance BrowseComp is an evaluation designed to test how well models can find hard-to-locate information on the web…
LLMs and biorisk
…As we noted at the time, this was a precautionary decision—improving model performance on our evaluations meant we could no longer confidently rule out the ability of our most advanced model…
Claude for Financial Services
…benchmarking and portfolio deep dives, financial modeling with full audit trails, and generating institutional-quality investment memos and pitch decks. Teams can monitor portfolio performance and compare metrics across investments to identify…
Golden Gate Claude
…models to review its systems, find vulnerabilities, and fix them. More details on Fable 5’s cyber safeguards and our jailbreak framework Introducing Claude Sonnet 5 Sonnet 5 delivers frontier performance across…
AI agents find smart contract exploits
…In the revenue vs models charts, we only show models that solved at least one problem. [4] This is according to each model's Best@8 performance. Best@8 means that we…
Anthropic Economic Index report: Learning curves
…Model selection The different Claude model classes (Haiku, Sonnet, and Opus) offer tradeoffs in terms of cost, speed, and performance. The Opus class of models uses the most tokens and excels at…