Evaluations
PreviewKnow a model is better before you ship it.
Score base models and fine-tunes on standard suites or your own tasks, and compare them in one matrix before you ship.
- Pricing
- Billed per GPU-hour
- Layer
- Training
- Compare
- Base models, Fine-tunes, Model checkpoints, Your datasets, Custom graders, Pipelines
Why Evaluations
Improvements are measured, not assumed.
Score base models and fine-tunes on standard reasoning, coding and tool-use suites, or on your own tasks with a dataset and a grader. Every run can be repeated, so a comparison still means something after the next fine-tune.
Standard suites
Reasoning, coding and tool-use benchmarks.
Custom tasks
Bring a dataset and a grader.
Comparison matrix
Every model against every suite, side by side.
How it works
Benchmark suites and model comparisons you can rerun.
In the console
Everything in one place.
Evaluations lives in the LegionEdge cloud console, next to every other service in the workspace, under one account and one identity.
Pricing
Billed per GPU-hour
Evaluation runs are billed per GPU-hour, at compute rates, for the GPUs they use.
- Metered the same way as fine-tuning.
- A pipeline run records what each of its steps cost.
GPU rates
Related products