Independent reference. We are independent of every vendor listed. No affiliate links. No sponsored placements.
AI guardrails

AI Guardrails Pricing 2026: Runtime LLM Guardrail Vendor Catalogue

Every runtime LLM guardrail vendor we track, grouped by pricing model. Hyperscaler-native is cheapest; specialists are quote-only; open-source is free at license.

9
Vendors with public price
$0.08
Cheapest per-1K paid rate
$0
Free OSS path floor
3
Hyperscalers with published rate

Hyperscaler-native (cheapest published)

B
Published
AWS Bedrock Guardrails
$0.08 per 1K text units (prompt attack filter)
→
M
Published
Google Cloud Model Armor
$0.10 per million tokens (after 2M free monthly)
→
A
Free tier
Azure AI Content Safety / Prompt Shields
Free F0 tier (5 RPS), then $0.38 per 1K text records on S0
→

Specialist runtime guardrails

L
Free tier
Lakera (Lakera Guard, Lakera Red)
Free Community tier, paid plans quote-only
→
G
Free tier
Guardrails AI
Open-source Hub free, Pro is quote-only
→
P
Quote only
Prompt Security
Quote only
→
C
Quote only
CalypsoAI
Quote only
→
P
Quote only
Pillar Security
Quote only
→
K
Quote only
Knostic
Quote only
→

Open-source (free at license)

N
Free tier
NVIDIA NeMo Guardrails
Free open source (Apache 2.0)
→
G
Free tier
Garak (NVIDIA)
Free open source (Apache 2.0)
→
W
Free tier[DISCONTINUED]
WhyLabs (LangKit)
Open source only (platform discontinued)
→

How to read pricing model

  • Per-1K-text-units / per-1K-text-records: pay-per-call. Linear with traffic.
  • Per-million-tokens: also pay-per-call, denominated in tokens (~4 chars each).
  • Flat-monthly: subscription. Usually with a usage cap.
  • Quote-only: no public price. Procurement-led, contract review required.
  • Open-source-free: zero license cost. FTE and infra remain on the buyer.

See /how-vendors-price-llm-security for the full pricing-model taxonomy.

Last verified June 2026Source: vendor pricing pages. See /methodology.