A structured open dataset mapping language model benchmark scores, context window limits, and provider pricing.
Access structured JSON datasets comparing performance metrics, provider costs, and model capabilities.
This repository provides a structured JSON dataset for evaluating large language models. Content strategists and marketers can inspect model context windows, performance scores, and API pricing to select cost-effective models. Note that this project is officially deprecated and no longer receives active updates.
It aggregates performance benchmark results (such as MMLU, GPQA, and MATH), parameter counts, context limits, and vendor pricing inside schema-validated JSON files. The dataset organizes records across dedicated directories for models, providers, and benchmarks. You can inspect these files directly to build custom comparison tools or analyze model efficiency.
It suits technical content creators, prompt engineers, and marketing operations specialists who need raw historical data to compare model performance against API costs. It does not suit non-technical users seeking a turnkey SaaS platform or teams requiring live, auto-updating benchmark feeds.
Model metadata, benchmark results, and pricing details are stored in clean, schema-validated JSON directories.
Links specific language models to provider pricing, context window parameters, and hardware configurations.
Records model scores across established industry benchmarks including MMLU, GPQA, HumanEval, and MATH.
Data entry schemas encourage verifiable citation links for all submitted performance metrics.
The repository content is free to download and inspect. The only cost involved is the time and hosting infrastructure required to run your own custom parsing scripts or web viewers.
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
No, the repository is officially deprecated. Data corrections are now directed to the llm-stats community section.
You can compare context window sizes and pricing against benchmark scores to pick the right LLM for long-form generation.
See Zekai first in Google