AIJev reportedly matches GPT-5.6 Luna’s reranking accuracy in Ramp benchmark
A user reports 10x lower tail latency, down to 300ms, at 3x lower cost in testing for Ramp’s accounting product.
TLDR
A user reports that Jev was benchmarked as a replacement for LLM-based reranking in Ramp’s accounting product. They say it matched current accuracy on GPT-5.6 Luna while reducing tail latency 10x to 300ms at 3x lower cost. On September 24, 2026, the author said they wanted to bring Jev into production for 70,000 customers but were still running into rate limits.
Jev reportedly matches GPT-5.6 Luna’s reranking accuracy in Ramp benchmark
A user reports 10x lower tail latency, down to 300ms, at 3x lower cost in testing for Ramp’s accounting product.
