Benchmarks seem like a fools errand at this point; overly tuning models just to specific test already published tests, rather than focusing on making them generalize.

Hugging face has a leader board and it seems dominated by models that are finetunings of various common open source models, yet don't seem be broader used:

https://huggingface.co/open-llm-leaderboard

Why is this sponsored by such an American gov entity?

I guess it's one of those things you never think about until it goes wrong.

The world would do well to move this kind of stuff out of the US quickly, just like ICANN and stuff.

It’s a reckless move to cut funding so abruptly, but taking a step back from the short-term chaos, it probably is an anomaly that this was government funded. All of private tech relies on it, and private tech is big enough to pay for it. I hope that the trillion dollar babies consider this an opportunity to pool together to form a foundation that funds this, and a bunch of other open source projects run by one random person in Nebraska.

Just curious, as I'm in the market - why should I use this instead of the ELK stack?

2.5 GbE is too slow for my home lab. I want 10 GbE-T PoE++ without having to add a craptastic hat that has a lot of failures.

My network test boxes run 100 GbE of various flavors supported by E810-CQDA2.