Custom Models

Register fine-tuned models, new models, or custom deployments.

Quick Start

const client = new LLMist({
  customModels: [{
    provider: 'openai',
    modelId: 'ft:gpt-5:my-org:custom:abc123',
    displayName: 'My Fine-tuned GPT-5',
    contextWindow: 272_000,
    maxOutputTokens: 128_000,
    pricing: { input: 2.5, output: 20.0 },
    knowledgeCutoff: '2024-09',
    features: {
      streaming: true,
      functionCalling: true,
      vision: true,
    },
  }],
});

Register at Runtime

const client = new LLMist();

client.modelRegistry.registerModel({
  provider: 'openai',
  modelId: 'gpt-5-preview',
  displayName: 'GPT-5 Preview',
  contextWindow: 200_000,
  maxOutputTokens: 32_000,
  pricing: { input: 15.0, output: 60.0 },
  knowledgeCutoff: '2025-01',
  features: {
    streaming: true,
    functionCalling: true,
    vision: true,
    reasoning: true,
  },
});

.withModel('openai:gpt-5-preview')

ModelSpec Reference

Field	Type	Required	Description
provider	string	Yes	Provider ID
modelId	string	Yes	Full model identifier
displayName	string	Yes	Human-readable name
contextWindow	number	Yes	Max context tokens
maxOutputTokens	number	Yes	Max output tokens
pricing.input	number	Yes	Cost per 1M input tokens
pricing.output	number	Yes	Cost per 1M output tokens
knowledgeCutoff	string	Yes	Knowledge cutoff date
features.streaming	boolean	Yes	Supports streaming
features.functionCalling	boolean	Yes	Supports tools
features.vision	boolean	Yes	Supports images

See Also

• Model Catalog - Query models and costs
• Providers Guide - Provider configuration