What it measures
Enterprise information tasks test retrieval and reasoning over business content in a model-launch evaluation.
How to read the result
Look for whether the answer is supported by the relevant documents and whether important qualifiers survive synthesis.
Limits and comparison traps
The available source describes a deployment-oriented evaluation; source access, indexing, and permissions can dominate differences between complete systems.
Published evaluation
Weighted rubric accuracy copied from the exact data underlying the publisher chart. This is rubric credit rather than a percentage of fully completed workflows. Prompt, tool and grading details remain those of the release comparison.
Where this fits in an evaluation suite
A large context window is a capacity specification. Successful retrieval and cross-document reasoning are observed capabilities. Test both at the actual lengths your workflow needs.
Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.
A reproducible comparison
Weighted rubric accuracy copied from the exact data underlying the publisher chart. This is rubric credit rather than a percentage of fully completed workflows. Prompt, tool and grading details remain those of the release comparison.
Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.
PRIMARY EVIDENCE & FURTHER READING
Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.