RightNow launches inference API, claims 14x NVIDIA performance via GPU kernels

ycombinator · x · 2026-11-17

Y Combinator-backed RightNow AI launched its own inference API built on GPU kernel-level optimizations, claiming half the open-weight inference market already uses its kernels and that NVIDIA saw 14x performance gains with them.

The API offers pay-per-token pricing, batch at half price, reserved capacity, and dedicated deployments with zero data retention by default. Listed models include DeepSeek V4.1 Flash ($0.15/M input, $0.60/M output, 327 tok/s), GLM 5.3 and GLM 5.3 Flash, and NVIDIA Nemotron 3.5 Lightning 30B (540 tok/s), with cache hit rates above 93%.

Original post →

More from Venture

Venture channel →