Language Models

Language models handle text generation and chat: single-prompt completions, multi-turn conversations, reasoning, and tool use. They are the most common model type on Gateway.

When to Use

Use a language model when you need to generate or transform text:

• Chat assistants and multi-turn conversations
• Single-prompt generation, summarization, and rewriting
• Reasoning tasks (use a reasoning-capable model with reasoningEffort)
• Code generation and analysis
• Tool calling and structured output

For image input, see Vision. For turning speech into text, see Audio.

AI SDK Method

Language models are accessed with chatModel() and used with the AI SDK's generateText and streamText:

import { createOLLM } from '@orgn/gateway';
import { generateText } from 'ai';

const ollm = createOLLM({ apiKey: process.env.OLLM_API_KEY });

const { text } = await generateText({
  model: ollm.chatModel('near_glm_5_1'),
  prompt: 'What is ORGN Gateway?',
});

For streaming, system messages, multi-turn conversations, and reasoning options, see the Vercel AI SDK integration.

TEE Catalog

Language models running in Trusted Execution Environments, on NEAR and Phala infrastructure with Intel TDX + NVIDIA H100 confidential compute, or on Tinfoil infrastructure with AMD SEV-SNP + NVIDIA confidential compute. Every request produces a cryptographic attestation receipt.

ZDR Catalog

Language models running on Vercel's AI infrastructure with zero data retention provider agreements. No attestation receipts are generated.

Anthropic

OpenAI

Google

xAI

Meta

Mistral

Alibaba (Qwen)

DeepSeek

Moonshot

ZAI

Other Language Models

Not confidential

A small set of catalog entries run through Vercel's infrastructure but carry neither a TEE attestation nor a ZDR agreement. See Models overview for what this tier does and does not guarantee.