specification: API Commons Plans
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/Plans
provider: Predibase
providerId: predibase
created: '2026-06-20'
modified: '2026-06-20'
reconciled: true
tags:
- AI
- LLM
- Fine-Tuning
- Inference
- LoRA
- Plans
description: >-
  Predibase uses usage-based pricing across three meters: serverless inference
  billed per token, fine-tuning (training) billed per token of training data
  scaled by base-model size, and dedicated deployments billed per GPU-hour by
  accelerator type. A free tier provides serverless inference up to a daily and
  monthly token cap; the Developer tier adds self-serve dedicated A10/A100
  deployments; the Enterprise tier adds VPC / multi-GPU (A100/H100) deployments
  and negotiated terms.
notes: >-
  Representative rates as published on the Predibase pricing page and docs;
  fine-tuning per-token rates scale with model size and dedicated GPU-hour
  rates vary by accelerator. Verify current rates on the pricing page. Predibase
  was acquired by Rubrik in June 2025.
sources:
- https://predibase.com/pricing
- https://docs.predibase.com/user-guide/inference/dedicated_deployments
- https://docs.predibase.com/user-guide/fine-tuning/overview
plans:
- id: predibase-free
  name: Free
  type: free
  description: >-
    Serverless inference on shared endpoints up to a daily and monthly token
    cap, with no cost to start.
  entries:
  - label: Serverless Inference (free allowance)
    name: serverless_free
    type: free
    metric: tokens
    limit: 10000000
    timeFrame: month
    geo: global
    unit: 1000000
    price: free up to ~1M tokens/day and ~10M tokens/month
    userMultiplied: false
  elements:
  - name: Shared Endpoints
  - name: OpenAI-Compatible Inference
- id: predibase-pay-as-you-go
  name: Pay-as-you-go
  type: usage
  description: >-
    Token-metered serverless inference, token-metered fine-tuning, and
    GPU-hour-metered dedicated deployments with no monthly minimum beyond usage.
  entries:
  - label: Serverless Inference
    name: serverless_inference
    type: usage
    metric: tokens
    limit: -1
    timeFrame: month
    geo: global
    unit: 1000000
    price: ~$0.20 per 1M tokens (small models; varies by model size)
    userMultiplied: false
  - label: Batch Inference
    name: batch_inference
    type: usage
    metric: tokens
    limit: -1
    timeFrame: month
    geo: global
    unit: 1000000
    price: ~$0.50 per 1M tokens (flat, input and output)
    userMultiplied: false
  - label: Fine-Tuning (up to 7B)
    name: finetuning_7b
    type: usage
    metric: tokens
    limit: -1
    timeFrame: month
    geo: global
    unit: 1000000
    price: from ~$0.36 per 1M training tokens
    userMultiplied: false
  - label: Fine-Tuning (large / MoE, e.g. Mixtral-8x7B)
    name: finetuning_large
    type: usage
    metric: tokens
    limit: -1
    timeFrame: month
    geo: global
    unit: 1000000
    price: up to ~$3.21 per 1M training tokens
    userMultiplied: false
  elements:
  - name: Serverless Inference
  - name: Batch Inference
  - name: Supervised Fine-Tuning
  - name: Reinforcement Fine-Tuning (GRPO)
- id: predibase-developer
  name: Developer (Dedicated Deployments)
  type: usage
  description: >-
    Self-serve dedicated deployments billed per GPU-hour on A10 and A100
    accelerators for production traffic.
  entries:
  - label: Dedicated A10G (24GB)
    name: dedicated_a10
    type: usage
    metric: hours
    limit: -1
    timeFrame: month
    geo: global
    unit: 1
    price: from ~$1.82 per GPU-hour
    userMultiplied: false
  - label: Dedicated A100 (80GB)
    name: dedicated_a100
    type: usage
    metric: hours
    limit: -1
    timeFrame: month
    geo: global
    unit: 1
    price: per GPU-hour (see pricing page)
    userMultiplied: false
  elements:
  - name: Dedicated Deployments
  - name: LoRA / Turbo LoRA Serving
- id: predibase-enterprise
  name: Enterprise
  type: enterprise
  description: >-
    VPC / private cloud deployments, multi-GPU and H100 deployments, dedicated
    support, and negotiated terms. Contact Predibase sales.
  entries:
  - label: Enterprise Agreement
    name: enterprise
    type: flat
    metric: contract
    limit: -1
    timeFrame: year
    geo: global
    unit: 1
    price: contact sales
    userMultiplied: false
  elements:
  - name: VPC / Private Deployments
  - name: H100 and Multi-GPU
  - name: Custom Volume Pricing
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com