The new Jev AI model utilizes a non-generative architecture to answer structured questions in parallel instead of producing text token by token. TypeSafe AI granted access to 140,000 people on its waitlist in less than 36 hours following the release of the tool, which provides typed choices and confidence scores. The system is currently available through the Vercel AI Gateway for developers building agents that require high-speed routing or scoring.
Deployment data show a 71% reduction in costs and a 90% reduction in completion time for a production dispatch process. Another test used Jev to classify 1,018 research papers for $0.08 with a 256 millisecond median latency per paper. TypeSafe AI reports that Jev performs 20 to 200 times faster and 40 to 400 times cheaper than text-generating models by removing the sequential token layer.
Key sources
- SOURCE@completeskeptic“we've jev-ed 140k off the waitlist (our platform team trying hard to get EVERYONE off ASAP)”x.com
- SUPPORT@kunchenguid“the saving from Jev still resulted in a -71% reduction in cost and -90% reduction in wall time of completing the whole dispatching process”x.com
- SUPPORT@completeskeptic“I used Jev to classify 1,018 AI research papers. The result: $0.08 total cost and 256ms median end-to-end latency per paper”x.com
- SUPPORT@completeskeptic“results of using jev on one of our security pipelines. we already have to use smaller (and dumber) models”x.com
- SOURCE@ctatedev“Input $0.04/M”x.com
- SUPPORT@rauchg“Jev is up to 18x faster (p95) *and* more accurate. It's coming to @vercel AI Gateway and likely new default.”x.com
- SUPPORT@rauchg“Jev is now available on the @vercel AI Gateway”x.com
- SOURCEmarketbrief.now