---
type: "page"
slug: "benchmarks"
title: "Benchmarks"
summary: "A transparent workbench for decision-specific model evaluation, with no winner before a real run."
creator: "Rahil Chamola"
updated: "2026-07-23"
canonical_url: "https://rahilchamola.com/benchmarks"
preferred_citation: "Chamola, Rahil. “Benchmarks.” Curated Randomness, 2026."
---

# Benchmarks

A transparent workbench for decision-specific model evaluation, with no winner before a real run.

## Current state

There are zero completed public benchmark runs. Proposed case and repeat counts are experiment designs, not executed calls or rankings.

## Evidence contract

A public result requires exact model identities, immutable inputs, prompt and dataset hashes, current pricing sources, captured failures, human calibration, raw artifacts, and a bounded decision rule.

---

Creator: Rahil Chamola
Canonical source: https://rahilchamola.com/benchmarks
Preferred citation: Chamola, Rahil. “Benchmarks.” Curated Randomness, 2026.