👉 GnollBench is the Gnoll Overseer's own AI benchmark. Before an AI model is offered in the Overseer, it has to pass GnollBench: a fixed exam of GnollHack questions, taken with the same instructions and lookup tools as the real chat.
📊 What GnollBench Is For
GnollBench exists first of all to find and fix weaknesses in the Overseer itself, not merely to rank one AI model above another. Every run reports three results — Intelligence, Speed, and Cost — so a model's strengths and its price are visible side by side.
📖 Guides
| Guide | For whom | What you get |
|---|---|---|
| Introduction to GnollBench | Newcomers | What GnollBench is and why it exists, in plain language |
| Advanced Guide to GnollBench | Interested readers | How scoring, grading, and repeated runs work |
| Technological Overview of GnollBench | Readers who know AI evaluation | Formulas, statistics, and comparability rules |
| Comparison of GnollBench and Popular AI Benchmarks | Anyone comparing benchmarks | How GnollBench compares with popular public AI benchmarks, and how they complement each other |
💡 Tip: Not sure where to start? Read the introduction first. Each GnollBench guide links to the next level of depth.
📖 See Also
- Gnoll Overseer — All Gnoll Overseer guides in one place