VRAM VM REPORT — 2026-07-02 08:24 UTC
Jane fingers · scrape → analyze · brain reads this

LAW: 24GB class · one pipe hot · stop pod when idle · NOT 160GB 24/7

TOP PICKS (SaaS · verify price before buy):
  1. Vast.ai RTX 3090 — $0.18/hr · ~$7.2/mo @40h drain · score 85
     Lowest 24GB class often · verify host reliability · 24GB forge fit · cheap 24GB · drain-friendly
  2. RunPod RTX 3090 — $0.22/hr · ~$8.8/mo @40h drain · score 55
     Cheaper 24GB · same forge stack law · 24GB forge fit · cheap 24GB · 24/7 breaks cap · drain-friendly
  3. RunPod RTX 4090 — $0.44/hr · ~$17.6/mo @40h drain · score 45
     Community cloud · stop when idle · good Forge drain lane · 24GB forge fit · 24/7 breaks cap · drain-friendly

FULL RANK:
  [PASS] vast-3090: $0.18/hr · 24GB · score 85
  [PASS] runpod-3090: $0.22/hr · 24GB · score 55
  [PASS] runpod-4090: $0.44/hr · 24GB · score 45
  [PASS] vast-4090: $0.35/hr · 24GB · score 45
  [PASS] lambda-a10: $0.75/hr · 24GB · score 10
  [PASS] paperspace-a5000: $0.78/hr · 24GB · score 10
  [CPU] linode-cpu: $0.03/hr · 0GB · score 5
  [SKIP] always-on-160gb: $1.65/hr · 160GB · score -100

BRAIN SAYS: RunPod/Vast 3090 or 4090 on-demand · stop when done · Linode stays voice home.
Word: vram shop · re-run after price check