62 / 100

Baseten pricing teardown

https://www.baseten.co/pricing

Baseten's pricing page is a usage-based, infrastructure-first page targeting ML engineers and AI teams who want transparent GPU/token pricing with a free entry point. The three-tier plan structure (Basic free, Pro quote-based, Enterprise quote-based) is supplemented by a detailed public rate card for Model APIs and Dedicated GPU/CPU compute, which is genuinely differentiated and trust-building. However, the Pro tier has no public price and both paid tiers require 'Get a quote,' creating friction for self-serve buyers who want to upgrade without a sales call.

Tier structure

Basic

$0/mo (pay-as-you-go usage)

Pro

Volume discounts, quote required

Enterprise

Volume discounts, quote required

Value metric

usage-based (per GPU-minute for compute, per 1M tokens for Model APIs)

How limits scale

Deployment options Baseten cloud only → Baseten cloud + dedicated compute → Baseten + Your VPC + Hybrid
GPU access Standard access → Priority access to high-demand GPUs → On-demand flex compute + custom global regions
Support Email and in-app chat → Dedicated Slack and Zoom + hands-on engineering → Custom SLAs + forward-deployed engineers
Model API rate limits Standard limits → Higher rate limits → Not specified (Enterprise custom)
Data residency / security SOC 2 Type II + HIPAA → SOC 2 Type II + HIPAA → Full data residency control + Advanced RBAC + Teams

Escalation logic

Basic is a free-to-start tier with pay-as-you-go compute access and core features including SOC 2/HIPAA compliance. Pro and Enterprise layer on priority GPU access, dedicated support, and (for Enterprise) VPC/self-host options and advanced security, but both require a sales quote rather than self-serve checkout.

Psychological anchors

  • Free entry anchor — Basic at $0/month with no credit card required lowers the barrier to entry and gets engineers into the product before any spend commitment.
  • Public rate card as trust signal — Detailed per-minute GPU pricing (T4 through B200) and per-1M-token Model API pricing are fully public, anchoring buyers on transparent unit economics before they talk to sales.
  • Enterprise as high-anchor tier — Enterprise lists VPC, hybrid, custom SLAs, and advanced RBAC to make Pro feel like the rational middle choice, even though Pro has no public price.
  • Volume discount mention — 'Volume discounts available' appears on both Pro and Enterprise tiers and in the compute rate card, implying savings at scale without committing to a number — a soft urgency nudge to contact sales.
  • Dual CTAs on hero — 'Start building' (self-serve) and 'Talk to an engineer' (sales) are paired at the top, segmenting intent immediately.
  • Free credits mention in FAQ — FAQ confirms new accounts get free credits, reinforcing the no-risk entry point without prominently featuring it in the plan cards.

What this page is optimizing for

Baseten is optimizing for a hybrid of self-serve developer acquisition (free Basic tier, public rate card, instant 'Deploy' and 'Try Model API' CTAs) and enterprise lead generation (Pro and Enterprise both gate behind 'Get a quote'). The page is clearly built to get engineers experimenting immediately while routing higher-value accounts to sales.

Red flags

  • Pro tier has no public price, forcing even mid-market buyers into a sales conversation that may deter self-serve-oriented engineering teams.
  • No annual vs. monthly toggle is present, missing a standard conversion lever and discount signal that most SaaS buyers expect.
  • The plan feature comparison is thin — differences between Basic and Pro are vague (e.g., 'higher rate limits' with no numbers), making upgrade triggers unclear.
  • No 'most popular' or recommended plan badge is detectable from the text, weakening the decoy effect and leaving buyers without a clear default choice.
  • Free credits are only mentioned in the FAQ, not in the Basic plan card itself — a missed opportunity to reduce signup hesitation at the point of decision.
  • The page mixes plan tiers, a model API catalog, and a GPU rate card in one scroll, which may overwhelm buyers who just want to understand plan differences.
  • No social proof, customer logos, or usage statistics are visible on the pricing page itself to build trust at the moment of purchase intent.

Best-practices scorecard

What works · 2

  • Value metric matches usage — Charging per GPU-minute and per token is well-aligned with how ML engineers actually consume inference infrastructure.
  • FAQ or objection handling — Eight FAQ items address idle billing, security, GPU availability, free credits, and self-hosting — covering the most common objections.

Half measures · 4

  • Visible prices — Basic ($0) and the full compute/token rate card are public, but Pro and Enterprise prices are hidden behind 'Get a quote.'
  • Tier differences are scannable — Feature lists exist per tier but key limits (rate limits, compute quotas) lack specific numbers, making comparison difficult.
  • Trust signals present — SOC 2 Type II and HIPAA compliance are mentioned in the Basic plan card and FAQ, but no customer logos or case studies appear on the pricing page.
  • Clear CTAs per tier — Basic has 'Get started' and compute rows have 'Deploy' CTAs, but Pro and Enterprise only offer 'Get a quote' with no self-serve path.

What's missing · 2

  • Clear recommended tier — No 'most popular' or highlighted plan is detectable in the text; buyers have no guided default.
  • Annual discount offered — No annual billing toggle or discount is mentioned anywhere on the page.