Benchmark and arena scores per model, with each model's prices in DA.
GET /v1/benchmarks returns benchmark and arena scores per model. It is free and takes an API key. The openai SDK has no benchmarks resource, so call it with fetch.
const params = new URLSearchParams({ task_type: "coding", max_results: "5" });
const res = await fetch(`https://opendunes.com/v1/benchmarks?${params}`, {
headers: { Authorization: `Bearer ${process.env.OPENDUNES_API_KEY}` },
});
if (!res.ok) throw new Error(`benchmarks: ${res.status}`);
const { data } = await res.json();
for (const row of data) console.log(row.source, row.display_name, row.coding_index, row.pricing);
// pricing is the model's price in DA per token| Parameter | Takes |
|---|---|
source | artificial-analysis · design-arena · opendunes |
task_type | coding · intelligence · agentic · search |
benchmark_type | a single benchmark, such as gpqa_diamond |
arena · category | a design arena and one of its categories |
max_results | how many rows to return |
A value the route does not take is a 400 naming the values it does. When the scores cannot be read, the route answers 503 with a Retry-After.