Finally, an LLM benchmark that can’t be faked

I’m pretty sure this was Microsoft’s versioning logic for the 90s and 2000s.

Ok! Time to relax. Sashi is a shitposter.

https://x.com/sashikantsingh_/status/2105531543666217357

From the Line Goes Up school of thought.

thatsthejoke.tensor

Grok 86.47 for the win!

Sounds reliable. There’s a name for claims like that https://github.com/ajnart/trustmebrobenchmarks.com