aid: parasail
specVersion: '0.1'
name: Parasail Plans & Pricing
url: https://www.saas.parasail.io/pricing
description: |
  Parasail offers four commercial surfaces — Serverless, Dedicated Serverless,
  Dedicated, and Batch — billed on a pay-per-token or GPU-hour basis with no long-term
  contracts. Tiers (Free, User, Dedicated Serverless, Dedicated Serverless Pro,
  Enterprise) gate request-per-minute capacity. Free credits are provided for new
  accounts.
plans:
  - id: free
    name: Free
    description: Free tier with starter credits for evaluating Parasail's serverless inference.
    entries:
      - geo: Global
        unit: 1
        label: User
        limit: 1
        price: 0
        metric: user
        timeFrame: month
        description: Free starter credits, capped at 5 RPM.
    elements:
      - name: Pay-per-token serverless inference (after free credits exhausted)
      - name: Access to all serverless models exposed on /v1/models
      - name: OpenAI-compatible /v1/chat/completions, /v1/completions, /v1/embeddings
      - name: 5 RPM rate limit
      - name: Free credits for new users
  - id: user
    name: User
    description: Standard pay-per-token serverless tier for individual developers and small teams.
    entries:
      - geo: Global
        unit: 1
        label: User
        price: PayPerToken
        metric: user
        timeFrame: month
        description: Pay-per-token usage-based billing, 500 RPM.
    elements:
      - name: 500 RPM
      - name: All serverless models
      - name: Batch API access (50% off serverless, +30% off cached tokens)
      - name: No quotas on monthly token volume
  - id: dedicated-serverless
    name: Dedicated Serverless
    description: |
      Guaranteed throughput against a chosen model on isolated capacity, still billed
      per token but with reserved GPUs behind the endpoint.
    entries:
      - geo: Global
        unit: 1
        label: Deployment
        price: Reserved
        metric: deployment
        timeFrame: month
        description: Reserved isolated pool with per-token billing and 1,000 RPM ceiling.
    elements:
      - name: 1,000 RPM
      - name: Isolated capacity for a chosen model
      - name: Pay-per-token billing on reserved pool
      - name: Control-plane API for pause/resume/scale
  - id: dedicated-serverless-pro
    name: Dedicated Serverless Pro
    description: Higher-throughput dedicated serverless tier for production workloads.
    entries:
      - geo: Global
        unit: 1
        label: Deployment
        price: Reserved
        metric: deployment
        timeFrame: month
        description: Reserved isolated pool with 4,000 RPM ceiling.
    elements:
      - name: 4,000 RPM
      - name: Production-grade SLOs
      - name: All Dedicated Serverless features
  - id: dedicated
    name: Dedicated
    description: |
      Fully reserved GPU deployments billed on GPU-hours. Bring any Hugging Face or
      custom model and choose the device SKU and replica count.
    entries:
      - geo: Global
        unit: 1
        label: GPU
        price: GPUHour
        metric: gpu-hour
        timeFrame: hour
        description: Billed per GPU-hour against a chosen device config and replica count.
    elements:
      - name: Bring-your-own model (any Hugging Face / custom)
      - name: Choose GPU SKU (H100, A100, H200, etc.)
      - name: Autoscaling between min/max replicas
      - name: Pause and resume to control cost
  - id: batch
    name: Batch
    description: Asynchronous batch inference for offline workloads at 50% off serverless rates.
    entries:
      - geo: Global
        unit: 1M
        label: Tokens (Input + Output)
        price: 50PctOffServerless
        metric: token
        timeFrame: usage
        description: 50% off the corresponding serverless model rate, with cached tokens at an additional 30% off.
    elements:
      - name: 24-hour completion window
      - name: Supports /v1/chat/completions and /v1/embeddings
      - name: OpenAI-compatible JSONL batch file format
      - name: 80-90% cheaper than real-time for large offline jobs (combined with caching)
  - id: enterprise
    name: Enterprise
    description: Custom contracts for large-scale tokenmaxxing customers.
    entries:
      - geo: Global
        unit: 1
        label: Contract
        price: Call
        metric: contract
        timeFrame: year
        description: Custom pricing, unlimited RPM, dedicated support.
    elements:
      - name: Unlimited RPM
      - name: Custom model onboarding and dedicated capacity
      - name: Premium support and SLOs
      - name: Volume discounts