Model Benchmarks

W_S — Willingness to pay for speed (€ / token/s), log scale Min Max

W_I — Willingness to pay for intelligence (€ / point), log scale Min Max

Cost model Input:Output

Price = (ratio × input price) + output price. A higher ratio weights the input price more heavily (e.g. RAG with long context).

Open weights only

Models

Charts

Intelligence

Speed (tokens/s)

Price, split by input/output

Input Output

Intelligence vs. price (log scale)

Intelligence vs. speed (log scale)