> For the complete documentation index, see [llms.txt](https://docs.layerlens.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.layerlens.ai/5.-select-pick-the-model/quarterly-reports.md).

# Research — quarterly reports

Q1-Q4 research reports — model performance, benchmark winners, AI evaluation trends.

LayerLens publishes quarterly research reports that synthesize the platform's evaluation data into a public-facing read. Each report covers what's changed at the model frontier, what benchmarks revealed about specific capabilities, and where AI evaluation is heading.

## What's in a report

Each quarterly report includes:

* **Frontier model performance summary.** Which models advanced. Which plateaued. Where open-weight is closing the gap.
* **Benchmark deep-dives.** Two or three benchmarks of focus per quarter, with score deltas and surprising results.
* **Capability spotlights.** Reasoning, coding, agent tool-use, multilingual, multimodal — pick one or two and look at them through the leaderboard.
* **Methodology notes.** What changed in the platform's evaluation methodology this quarter, and why the change.
* **Trend forecasts.** Where the data suggests evaluation is heading next.

## Most recent: 2025 reports

Four quarterly reports were published in 2025, available on the public site:

* [**Q1 2025**](https://layerlens.ai/research/q1-2025) — frontier-model state at the top of the year
* [**Q2 2025**](https://layerlens.ai/research/q2-2025) — mid-year update
* [**Q3 2025**](https://layerlens.ai/research/q3-2025) — third-quarter snapshot
* [**Q4 2025**](https://layerlens.ai/research/q4-2025) — annual review

Index: [layerlens.ai/research](https://layerlens.ai/research).

## How to use a report

* **Choosing a model?** Read the most recent report's frontier summary, then jump into the [Compare models](/5.-select-pick-the-model/compare-models.md) feature with the candidates you're left with.
* **Building a business case?** Use the report's data to anchor "the model frontier moves quarterly — your evaluation pipeline should too."
* **Researching a benchmark?** Find the quarter that highlighted that benchmark; the report's methodology notes are the most rigorous public discussion you'll find.

## Where Stratix data comes from

The reports draw on the public catalog: 175+ models scored against 52+ benchmarks, refreshed continuously. Public evaluations on `stratix.layerlens.ai` are the underlying sample frame; the report adds analysis and narrative.

## Where to next

* [Public evaluations](/5.-select-pick-the-model/public-evaluations.md)
* [Models catalog](/5.-select-pick-the-model/models-catalog.md)
* [Benchmarks catalog](/5.-select-pick-the-model/benchmarks-catalog.md)
* [How Stratix compares](/1.-introduction/how-stratix-compares-deep.md)