Megadose AI progress, ranked and analyzed.

Show HN: JevBench, a reproducible benchmark for typed decision models

· HN · GitHub AI ·
classifier.dev is listed as an honorable mention because its fast tier is Jev 1.13.0, not a separate model.

JevBench says ranking classifier.dev against Jev would amount to ranking Jev’s model against itself. The benchmark measured only the fast tier, while the smart tier’s low-confidence escalation layer was not run or scored. Its $0.0033 per 1,000 decisions figure is an estimate based on full use of classifier.dev’s Pro plan, with lower usage raising the effective price. HN · GitHub AI's note

score 4

Categories: OSS & Tools, Research

Discussions

  • hn · 121 points · 32 comments
  • hn · 122 points · 33 comments
  • hn · 123 points · 33 comments