Compare models

Compare any two models head-to-head across any shared benchmark — Stratix Public's flagship feature.

The Compare models feature is Stratix Public's flagship: pick any two models, see them head-to-head across every benchmark either has been evaluated against.

URL: stratix.layerlens.ai/compare-models

What you can do

• Pick any two models from the catalog
• See a side-by-side score table across all shared benchmarks
• Drill into a single benchmark for a deeper comparison
• See where each model wins, ties, and loses
• Bookmark the comparison URL — it's stable

How it's structured

The compare-models page has three regions:

1. Selector — search and pick model A and model B
2. Score table — one row per benchmark, model-A score, model-B score, winner
3. Detail view — click any benchmark to expand: raw scores, sample inputs, sample outputs

How comparisons render

For each benchmark:

• Stronger model gets a green check + score
• Weaker model gets the score plain
• Tie is shown gray
• The gap (numeric and percentage) is shown

When the gap is small, the page shows a confidence band — was the difference statistically meaningful given the benchmark sample size?

Three or more models

The Premium Evaluations surface supports comparing many models on your own dataset; the public Compare-models page is intentionally pairwise (this matches how most users decide).

Common patterns

• Frontier vs cost-optimized — the same provider's flagship vs a smaller, cheaper sibling
• Provider against provider — OpenAI's flagship vs Anthropic's vs Google's
• Old vs new generation — when a new model drops, immediately compare against its predecessor

Where to next

• Premium — Evaluations — compare models on your data
• Models catalog
• Use case: Model evaluation