IBM watsonx.ai Lite Plan

Overview

IBM watsonx.ai is IBM's enterprise studio for foundation models and machine learning. The Lite plan is described in IBM's watsonx.ai Runtime service plans as a free plan with limited capacity, intended for evaluating and prototyping on the platform.

Free Tier Quotas & Limits

• 20 Capacity Unit Hours (CUH) per month for watsonx.ai Runtime (AutoAI training, model deployment and scoring)
• 300,000 foundation-model tokens per month for inferencing (1,000 tokens = 1 Resource Unit)
• 100 document text classification and extraction pages per month
• Rate limit: 2 inference requests per second
• Deployment time to idle: 1 day
• Maximum 2 parallel Decision Optimization batch jobs per deployment
• 1 watsonx.ai Studio instance per account

Key Free Models

Access to IBM's supported foundation model catalog, subject to capacity and data-center region, includes IBM Granite (granite-4-h-small, granite-8b-code-instruct, granite-guardian-3-8b), Meta Llama (llama-3-3-70b-instruct, llama-4-maverick-17b-128e-instruct-fp8), Mistral (mistral-small-3-1-24b-instruct-2503, mistral-large-2512), and OpenAI gpt-oss-120b.

Eligibility

One Lite plan instance per IBM Cloud account. Identity verification is required, but no charges apply unless you upgrade. The Lite plan does not support foundation model tuning, custom model deployments, GPU runtime environments, or project export.

How to Get Started

Register for watsonx.ai, choose an IBM Cloud region (Dallas or Frankfurt offer the widest model availability), and open the watsonx.ai studio. Use the Prompt Lab for interactive testing or the ibm-watsonx-ai SDK for programmatic access. Track CUH and token usage in the IBM Cloud dashboard.