typesafe-ai-benchmark

readme より

inference-benchmarks Benchmarks for model × version × engine. A score is never just the model's. The same weights score differently at another quant, or in another engine. So every result here names the exact checkpoint, quant and engine that produced it, and one harness measures them all the same way. Hackers in the Loop · runs on recipes from …