> ## Documentation Index > Fetch the complete documentation index at: https://docs.inference.net/llms.txt > Use this file to discover all available pages before exploring further. # Read the Results > Interpret the side-by-side comparison view and decide which model wins. After an eval completes, the comparison view shows how each model performed across every sample and rubric dimension. ## What the comparison shows * **Side-by-side plots** highlighting where models differ in quality * **Full scores table** across all models and samples * **Per-sample breakdown** so you can see where specific models excel or struggle Eval comparison view showing side-by-side model scores and the scores table. ## How to read the results Look for: * **Overall winner** - which model has the highest average score across your rubric * **Edge cases** - samples where one model significantly outperforms another * **Rubric dimensions** - if you have multiple rubrics, check whether models trade off on different quality dimensions (e.g. one model is more accurate but another has better tone) While the aggregate scores can inform you on how different models perform in a nutshell, it is recommended to analyze individual samples in the sample viewer. This will help to understand specific model quirks. ## Making decisions * **Which model to use in production** - the one that best matches your quality criteria * **Whether to train a custom model** - if no off-the-shelf model scores well enough, [fine-tuning](/platform/train/overview) is the next step * **Whether the rubric needs work** - if scores don't align with your intuition, iterate on the rubric before changing models