Hacker News AI· OsamaJaber·· 1 days agoAI score58
Show HN: Own your AI at full speed
Show HN: Own your AI at full speed
AI summary
A Show HN post presents a GPU-kernel-optimized inference service for open models, listing per-million-token prices and measured speeds for models including DeepSeek V4.1 Flash, GLM 5.3, Nemotron 3.5 Lightning 30B, Qwen3.8 27B and Ornith 1.5 35B. It describes automatic best-effort prompt caching, a five-layer serving stack, and pay-as-you-go, reserved, dedicated and batch capacity options, with reserved and dedicated not yet offered. The API accepts Chat Completions and Messages requests, and new accounts get $1 free.
Source: Hacker News AI · rightnowai.co