# Los Doritos Benchmarks

> Reproducible, transparent AI model comparisons. Evidence over opinions. No sponsored rankings.

## About

- Mission: https://losdoritos.com/about
- Methodology: https://losdoritos.com/methodology
- Editorial Policy: https://losdoritos.com/about#editorial-policy
- Publisher: https://infowebplus.com

## Benchmarks

- [Research Synthesis — Climate Policy Papers (2020–2024)](https://losdoritos.com/benchmarks/climate-policy-research-synthesis): Models synthesize findings from 8 provided research papers on carbon pricing. Scored on citation accuracy, synthesis quality, and hallucination rate.
- [Chart Data Extraction — Quarterly Revenue Bar Chart](https://losdoritos.com/benchmarks/chart-data-extraction-vision): Vision-capable models extract structured data from a bar chart image. Scored on numerical accuracy and schema compliance.
- [Technical Blog Post — Kubernetes Networking Explainer](https://losdoritos.com/benchmarks/kubernetes-networking-writing): Models write a 1,200-word technical blog post explaining Kubernetes networking to intermediate developers. Evaluated on accuracy, structure, and clarity.
- [Python Bug Fix — Null Reference in Async Handler](https://losdoritos.com/benchmarks/python-bug-fix-async-handler): Four frontier models attempt to fix a subtle async/await bug in a FastAPI endpoint. Evaluated on correctness, test pass rate, and minimal diff size.
- [Multi-Step Logic Puzzle — River Crossing with Constraints](https://losdoritos.com/benchmarks/river-crossing-reasoning): Five models solve a classic reasoning puzzle with added constraints. Scored on logical validity, step completeness, and constraint satisfaction.

## Categories

- [Coding](https://losdoritos.com/categories/coding): Code generation, debugging, and software engineering tasks
- [Reasoning](https://losdoritos.com/categories/reasoning): Logic, math, and multi-step problem solving
- [Writing](https://losdoritos.com/categories/writing): Clarity, tone, structure, and editorial quality
- [Business](https://losdoritos.com/categories/business): Analysis, strategy, and professional communication
- [SEO](https://losdoritos.com/categories/seo): Search optimization, metadata, and content structure
- [Research](https://losdoritos.com/categories/research): Literature synthesis, citations, and fact-finding
- [Agents](https://losdoritos.com/categories/agents): Tool use, planning, and autonomous workflows
- [Vision](https://losdoritos.com/categories/vision): Image understanding, OCR, and visual reasoning
- [Math](https://losdoritos.com/categories/math): Symbolic math, proofs, and quantitative accuracy
- [Translation](https://losdoritos.com/categories/translation): Multilingual fidelity and nuance preservation
- [Automation](https://losdoritos.com/categories/automation): Scripting, APIs, and workflow execution
- [Enterprise](https://losdoritos.com/categories/enterprise): Compliance, security, and production readiness

## Methodology Updates

- [Prompt Standardization Guidelines](https://losdoritos.com/methodology/updates#prompt-standardization): How we standardize prompts across benchmarks to ensure fair, reproducible comparisons.
- [Introducing Our Benchmark Scoring Rubric v1.0](https://losdoritos.com/methodology/updates#scoring-rubric-v1): How we score benchmarks: weighted criteria, human review protocols, and why we publish every rubric.

## Reports

- [Mid-Tier AI Models vs Enterprise Java Migrations](https://losdoritos.com/reports/mid-tier-llms-java-migration-scarfbench-benchmark): A report-driven benchmark proposal comparing Claude Sonnet 5, GPT-5.6 Terra/Sol, DeepSeek V4, and Gemini 3.5 Flash on ScarfBench-style Java migration tasks.

## Key Pages

- Leaderboard: https://losdoritos.com/leaderboard
- Methodology Updates: https://losdoritos.com/methodology/updates
- Reports: https://losdoritos.com/reports
- Search: https://losdoritos.com/search
- RSS: https://losdoritos.com/rss.xml

## Principles

- Evidence over opinions
- Reproducibility with published prompts and parameters
- No sponsored rankings
- No invented results
- CC BY 4.0 for benchmark data