tools

Grounded Language Model

Grounded Language Model generates answers from retrieved documents, adds inline source attributions, and reduces unsupported claims in RAG applications.

In inglese

Grounded Language Model is Contextual AI's model for retrieval-augmented generation and agentic applications. It prioritizes retrieved knowledge over pretrained knowledge, provides inline attributions, and can separate factual content from commentary.

Teams access it through the /generate API or Contextual AI agents. The service includes free credits for the first 1M input and 1M output tokens; additional usage is charged by token, and custom rate limits and pricing require contacting Contextual AI.

Features

Generates responses grounded in retrieved documents
Provides inline attributions to source documents
Prioritizes retrieved knowledge over pretrained knowledge
Supports an avoid_commentary control
Can refuse to answer when supplied documents are not relevant
Available through the standalone /generate API
Available as the default model for new agents

Use cases

Answer questions from enterprise knowledge bases
Build customer-support assistants grounded in current policies
Support engineering research over technical documentation
Create finance research tools that cite source material
Add grounded generation to an existing RAG pipeline

Pros

Cons

Capabilities

Developer API — “Documentation : /generate API , Python SDK , Langchain package – and code examples for each of these!” source

Get it

Documentation

Security

SOC 2 Type II — “Our commitment to security excellence is validated through successful completion of SOC 2 Type II audit.” source
SOC 2 — “Our commitment to security excellence is validated through successful completion of SOC 2 Type II audit.” source

Pricing

Prices checked: 2026-09-25

Generate

Input: $3 / 1M tokens
Output: $15 / 1M tokens

The most grounded large language model in the world, engineered specifically to minimize hallucinations

Official website