the benchmark is like a tycoon game/simulation, pretty fun and probably not benchmaxxed

all 5 comments

sorted by: hot top controversial new old
[–] 5 points 1 week ago (3 children)

Who serves Chinese out of a food truck?

... oh, nevermind.

I love their summaries of troubled runs, cycling on the front page. "8.9% fulfillment rate. In any other industry, this would trigger a congressional hearing."

  • source
  • hideshow 3 child comments
  • [–] [S] 4 points 1 week ago (2 children)

    I thought Kimi K3 was funny https://foodtruckbench.com/blog/kimi-k3

    It almost seems like it was trying to be as helpful as possible by feeding as many people as it could, too nice for its own good

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 1 week ago

    It's a good benchmark. 3.1 is worse than 2.5 on my language test. 3 got it right, but it might have been lucky. web agent is cool. It downloaded programming language and tried running test, but bad initial knowledge doomed it to failure as it blew through all token allowance on a 5 word prompt.

  • source