APEX-Agents: Results as Published

APEX-Agents tests whether an AI agent can carry out long-horizon, cross-application professional work. The tasks were written by investment banking analysts, management consultants and corporate lawyers, and the agent completes each one inside a simulated work environment with files and tools (documents, spreadsheets, PDFs, mail, chat, calendar and a file system), with web search disabled. The result is Pass@1, the first-attempt success rate. The original release has 480 tasks; version 1.1 (September 2026) has 80 tasks per profession, a judge that scores hedged multiple-answer items as zero, and updated environments.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.