About
Independent AI benchmark research.
Mission
Los Doritos Benchmarks exists to build one of the most trustworthy public collections of AI benchmark experiments. We believe the AI industry needs independent, reproducible evaluations — not marketing benchmarks designed to make specific models look good.
Our core principles:
- Evidence over opinions — Every claim is backed by documented experiments
- Reproducibility — Prompts, parameters, and environments are published
- Transparent methodology — Scoring rubrics and review protocols are public
- No sponsored rankings — No pay-to-win, no affiliate-driven comparisons
- No invented results — Every score comes from a documented test run
Editorial Policy
We publish benchmark experiments designed to answer specific, well-defined questions. We do not publish AI news, product announcements, or opinion pieces about model capabilities without experimental evidence.
Our methodology updates are limited to:
- Methodology updates and versioning
- Scoring rubric changes with changelogs
- Corrections to published benchmarks
When we discover errors in published benchmarks, we issue corrections with full transparency about what changed and why. We do not silently edit results.
Scoring Method
Each benchmark defines its own scoring rubric, published on the benchmark page. Common patterns:
- Outcome correctness: 50–70% of total weight
- Process quality (explanations, formatting): 10–30%
- Efficiency (cost, latency): reported always, weighted only when relevant
Human review, when used, follows a blind protocol with a minimum of 2 independent reviewers. See our Scoring Rubric v1.0 for details.
Disclosure
Los Doritos Benchmarks is an independent research project published by InfoWebPlus . We are not affiliated with any AI model provider, cloud platform, or venture capital firm.
- We pay for our own API access to test models
- We do not accept payment for rankings or featured placement
- We do not use affiliate links
- We do not accept free API credits in exchange for favorable coverage
If this ever changes, it will be disclosed on this page immediately.
FAQ
Do you accept sponsored benchmark placements?
No. We do not accept payment for rankings, featured placement, or favorable scoring. Our disclosure policy is published on this page.
Can I reproduce your benchmarks?
Yes. Every benchmark publishes exact prompts, parameters, and execution details. Downloadable artifacts (JSON, prompt packs) are provided where possible.
Do you publish AI news?
No. This is a benchmark platform, not a news site. Methodology changes are published in our methodology documentation and updates pages.