Kingy AI research
Evidence you can inspect, reuse, and cite.
Three public research products turn difficult AI questions into versioned evidence: what API prices actually cost, how model prices change over time, and what can be verified about AI distribution.
Published products
Start with the question you need answered.
Each release has a plain-language explanation, a downloadable chart, machine-readable files, a stable DOI, and a public source trail.
Compare API pricing
OpenAI and Anthropic API Price Index
Compare documented price records with the model, billing mode, source, and release details needed to make a fair comparison.
What it shows: 212 normalized records: 127 OpenAI and 85 Anthropic.
Verify: GitHub · Hugging Face · Zenodo
Track price changes
AI Model Price Stability
See what repeated archive captures can—and cannot—say about changes in publicly posted model prices.
What it shows: 31,946 retained observations across 1,861 archive captures; 23 are machine-detected candidates, not automatically verified changes.
Verify: GitHub · Hugging Face · Zenodo
Understand distribution evidence
State of AI Distribution
Learn what a retained evidence cohort can verify about identity, access, documentation, and unknown outcomes.
What it shows: a 25-record first cohort, with completeness tracked separately from quality or commercial outcome.
Use and reuse
Choose the format that fits your job.
Making a decision
Read the plain-language report first. The methodology and limitations tell you what the figures do not prove.
Doing analysis
Download the CSV or JSON, use the data dictionary, and keep the release version with your work.
Citing or sharing
Use the version DOI for an exact release and the concept DOI when you mean the evolving research product.
Checking the work
Follow the source records, checksums, code, and dataset cards on GitHub, Hugging Face, and Zenodo.
Research standards
Transparent about evidence, uncertainty, and independence.
Who made this
Research lead: Curtis Pyke, Kingy AI. Kingy AI controlled the questions, methods, analysis, conclusions, wording, and publication decisions. The work was independently funded: no covered vendor provided cash, in-kind consideration, prepublication review, or veto rights. Curtis Pyke declared no relevant financial or personal conflicts. AI systems assisted with packaging, code generation, and validation; Curtis Pyke reviewed the release decisions and accepts accountability.
Every product has its own methodology, limitations, license, correction policy, and stable identifiers. No product is a market census, a product endorsement, or a guarantee of future price or business outcomes.