Pricing

Baseten pricing context

Human-reviewed pricing summary paired with DevTune’s public AI search visibility benchmark.

Reviewed pricing summary

  • Baseten uses consumption-based pricing with no charges for idle time.
  • Dedicated Deployments are billed per compute minute by GPU instance type, ranging from T4 to NVIDIA B200/H100; customers configure autoscaling including scale-to-zero.
  • Model APIs are priced per million tokens (input + output), ranging approximately $0.20–$1.50/1M tokens depending on the model.
  • Three plan tiers exist: Basic (pay-as-you-go, free credits for new accounts), Pro (volume discounts negotiable), and Enterprise (custom pricing, self-hosted option, starting ~$5,000/month on AWS Marketplace).
  • Training jobs are billed per-minute on on-demand GPU compute.
  • Discounts on compute are negotiable under Pro and Enterprise plans.

Benchmark context

#6

of 10 in LLM Inference & Serverless GPU

1.3%

AI search visibility

Sources and verification

Pricing changes often. Treat this page as evaluation context and verify contract terms, usage limits, and add-ons against the vendor’s current materials before making a buying decision.