The LLM evaluation engine powering AI benchmarks and routers. We use multi-LLM evaluation for accurate and unbiased evaluation of LLM quality, cost and speed.
Joined June 2025
- Tweets59
- Following31
- Followers64
- Likes74
Autobench retweeted
Most commercial routers promise to cut your AI bill by 60%, 70%, even 80%. They don’t.
Finding 2 from our Router Benchmark: None of them beat a mid-tier model.
A router earns its keep only if it beats always calling Luna: better answers, and a lower bill. You do not need a router just to enter the prize zone. Luna is already there.
We checked 17 settings across 7 systems. None did both.
We bootstrapped the board 10,000 times. The chance any of them ever does is zero.
Most commercial routers claim they will save you 60%, 70%, even 80% of your bill. They don't.
We checked 17 settings across those 7 systems. A real router earns its keep only if it beats always calling Luna: better answers, and a lower bill. You do not need a router just to enter the prize zone. Luna is already there.
We bootstrapped the data 10,000 times. The chance any of them ever does is zero.
3/14
