Hacker News Reader

Which others?

There are several public benchmarks for decision models where they don't top every single one. https://benchmarkheaven.com/jev-models/all

I guess their cheeky "Warning: AI Benchmarks, don't trust them" is their way out of that one.

The way that site charges a fee[1] for benchmarking models, makes any benchmarks on it very sus, IMO. TBF, the benchmarking site is probably more profitable than TypeSafe, just like the joke about how WinRAR is more profitable than OpenAI[2]. :)

[1]: https://benchmarkheaven.com/submit

[2]: https://news.ycombinator.com/item?id=49797811

https://huggingface.co/spaces/multimodalart/jev-decision-ind... is also ranking just for the love of the game.

Given that benchmarks are useful, why not give them economic value?

There's nothing wrong with them wanting to capture some economic value, but the way they're monetizing it is a slippery slope of incentives.

OTOH, I trust the benchmarks from Apolinario's HF space that you've linked to.

Reply on news.ycombinator.com