Ofir Press: Good Benchmarks Must Be Hard, Verifiable, and Cared About—Cheap Is Out

Benchmark researcher Ofir Press says the hardest part of his job is creating criteria that are hard, verifiable, and cared about, adding that "cheap" has been mostly abandoned as good ideas get expensive.

2026-10-20 ~ 2026-10-20 · 2 related posts

• Ofir Press on benchmark-building: hard, verifiable, desirable — but 'cheap' is out — maxbittker · 2026-10-20
• Ofir Press: we've given up on 'cheap' as a good-benchmark criterion — OfirPress · 2026-10-20