Overview
Unified model-to-provider mapping registry that maps human-readable model names to their provider-specific identifiers, default parameters, and configuration details, powering the Helicone playground and prompt testing features.
Description
This file defines a modelMapping data structure organized by creator (OpenAI, Anthropic, Google, Meta, DeepSeek) that provides a comprehensive registry of LLM models. Each model entry contains default token costs, default parameters (e.g., max_tokens, stop, response_format), and an array of provider implementations mapping to provider-specific model strings. For example, "GPT-4o" maps to gpt-4o on OPENAI, gpt-4o on AZURE, and openai/gpt-4o-2024-11-20 on OPENROUTER.
The file also exports seven utility functions for querying the registry: looking up models by creator, finding providers for a model, building complete configurations with endpoints and body mappers, and reverse-looking up creators from provider model strings.
Usage
Use this module when working with the Helicone playground, prompt testing, or any feature that needs to present a unified view of available models across multiple providers. The utility functions allow retrieving model configurations including endpoint URLs, authentication headers, and request body mappers.
Code Reference
Source Location
Signature
import { getMapper } from "../../llm-mapper/path-mapper";
import { getProviderConfig } from "./providers";
import { Creator, CreatorModelMapping, Parameters, Provider } from "./types";
export const modelMapping: CreatorModelMapping = { ... };
export function getModelsForCreator(creator: Creator): string[];
export function getProvidersForModel(creator: Creator, modelName: string): Provider[];
export function getModelString(creator: Creator, modelName: string, provider: Provider): string | null;
export function getModelConfig(creator: Creator, modelName: string, provider: Provider): { ... } | null;
export function findCreatorForProviderAndModel(provider: Provider, modelString: string): Creator | null;
export function getModelsForProvider(provider: Provider): { creator: Creator; modelName: string; modelString: string }[];
export function getModelInfoFromModelString(modelString: string): { creator: Creator; modelName: string; provider: Provider } | null;
Import
import {
modelMapping,
getModelsForCreator,
getProvidersForModel,
getModelString,
getModelConfig,
findCreatorForProviderAndModel,
getModelsForProvider,
getModelInfoFromModelString,
} from "@helicone-package/cost/unified/models";
Model Registry
Creator: OpenAI
| Model Name |
Providers |
Default Parameters
|
| GPT-4o mini |
OPENAI, AZURE, OPENROUTER |
stop: [], response_format: true
|
| GPT-4o |
OPENAI, AZURE, OPENROUTER |
stop: [], response_format: true
|
| ChatGPT-4o |
OPENROUTER |
stop: [], response_format: true
|
| GPT-4.1 |
OPENAI, AZURE, OPENROUTER |
stop: [], response_format: true
|
| GPT-4.1 mini |
OPENAI, AZURE, OPENROUTER |
stop: [], response_format: true
|
| GPT-4.1 nano |
OPENAI, AZURE, OPENROUTER |
stop: [], response_format: true
|
| o3 mini |
OPENAI, AZURE, OPENROUTER |
reasoning_effort: "medium", stop: [], response_format: true
|
| o1 |
OPENAI, AZURE, OPENROUTER |
reasoning_effort: "medium", stop: [], response_format: true
|
Creator: Anthropic
| Model Name |
Providers |
Default Parameters
|
| Claude 3.7 Sonnet |
ANTHROPIC, BEDROCK, OPENROUTER |
max_tokens: 8192, stop: [], response_format: false
|
| Claude 3.5 Haiku |
ANTHROPIC, BEDROCK, OPENROUTER |
max_tokens: 8192, stop: [], response_format: false
|
| Claude 3.5 Sonnet |
ANTHROPIC, BEDROCK, OPENROUTER |
max_tokens: 8192, stop: [], response_format: false
|
| Claude 3 Opus |
ANTHROPIC, BEDROCK, OPENROUTER |
max_tokens: 4096, stop: [], response_format: false
|
| Claude Sonnet 4 |
ANTHROPIC, BEDROCK, OPENROUTER |
max_tokens: 8192, stop: [], response_format: false
|
Creator: Google
| Model Name |
Providers |
Default Parameters
|
| Gemini 2.0 Flash |
GOOGLE_GEMINI, GOOGLE_VERTEXAI, OPENROUTER |
response_format: true
|
| Gemini 2.0 Flash-Lite |
GOOGLE_GEMINI, GOOGLE_VERTEXAI, OPENROUTER |
response_format: true
|
| Gemini 2.0 Flash Thinking |
GOOGLE_GEMINI, GOOGLE_VERTEXAI, OPENROUTER |
response_format: true
|
| Gemini 2.0 Pro |
GOOGLE_GEMINI, GOOGLE_VERTEXAI |
response_format: true
|
| Gemini 1.5 Flash |
GOOGLE_GEMINI, GOOGLE_VERTEXAI, OPENROUTER |
response_format: true
|
| Gemini 1.5 Pro |
GOOGLE_GEMINI, GOOGLE_VERTEXAI, OPENROUTER |
response_format: true
|
Creator: Meta
| Model Name |
Providers |
Default Parameters
|
| Llama 3.3 70B Instruct |
AZURE, OPENROUTER |
response_format: false
|
| Llama 3.2 1B Instruct |
AZURE, OPENROUTER |
response_format: false
|
| Llama 3.2 3B Instruct |
AZURE, OPENROUTER |
response_format: false
|
| Llama 3.1 8B Instruct |
AZURE, OPENROUTER |
response_format: false
|
Creator: DeepSeek
| Model Name |
Providers |
Default Parameters
|
| DeepSeek V3 |
DEEPSEEK, OPENROUTER |
stop: [], response_format: false
|
| DeepSeek R1 |
DEEPSEEK, OPENROUTER |
stop: [], response_format: false
|
I/O Contract
Inputs (getModelConfig)
| Name |
Type |
Required |
Description
|
| creator |
Creator |
Yes |
Model creator (OpenAI, Anthropic, Google, Meta, DeepSeek)
|
| modelName |
string |
Yes |
Human-readable model name
|
| provider |
Provider |
Yes |
Target provider (OPENAI, ANTHROPIC, AZURE, BEDROCK, etc.)
|
Outputs (getModelConfig)
| Name |
Type |
Description
|
| provider |
Provider |
The resolved provider
|
| modelString |
string |
Provider-specific model identifier
|
| endpoint |
string |
API endpoint URL
|
| mapper |
function |
Request body mapper function
|
| baseUrl |
string |
Provider base URL
|
| authHeaderConfig |
object |
Authentication header configuration
|
| defaultHeaders |
object |
Default request headers
|
Usage Examples
import { getModelsForCreator, getModelConfig } from "@helicone-package/cost/unified/models";
// List all OpenAI models
const openaiModels = getModelsForCreator("OpenAI");
// ["GPT-4o mini", "GPT-4o", "ChatGPT-4o", "GPT-4.1", ...]
// Get config for GPT-4o on Azure
const config = getModelConfig("OpenAI", "GPT-4o", "AZURE");
// { provider: "AZURE", modelString: "gpt-4o", endpoint: "...", ... }
// Reverse lookup: find creator from model string
const creator = findCreatorForProviderAndModel("ANTHROPIC", "claude-3-7-sonnet-latest");
// "Anthropic"
Related Pages
