Price change

Cached-input pricing comes to Model APIs

Baseten pricing

Baseten adds discounted cached-input pricing to its multi-tenant Model APIs.

From Baseten's pricing timeline
Cached Input Pricing on Model APIs

Baseten added cached-input pricing to Model APIs — prompts re-using prefixes from prior requests within a session window are billed at 50–80% off the standard input rate (e.g., DeepSeek V3.1 input drops from $0.50 to $0.25 cached). This narrows the gap against first-party caching offerings from OpenAI and Anthropic.

About Baseten
baseten.co ↗

Baseten runs a pure-usage GPU-minute billing model for dedicated model deployments plus a separate per-token Model APIs catalog — both pay-as-you-go from the Basic tier with no monthly minimum.

Pricing model pure usagehybridcommitment
Billing units gpu hourstokensrequests
Sales motion self serveplgsales led
Free tier
Yes
Commits
Available
Transparency
public
Full Baseten breakdown Plans, meters, and history All pricing activity Track changes across AI companies

Baseten pricing history

  1. Feb 2026
    Cached Input Pricing on Model APIs
  2. Sep 2025
    Self-Hosted (BYOC) Deployment Option
  3. Mar 2025
    B200 GPU Availability + Mission Critical SLA Tier
  4. Nov 2024
    H100 GPU Pricing Published at $0.10833/min
  5. Aug 2024
    Model APIs (Multi-Tenant) Launched
Full Baseten timeline
All pricing activity