Model Mapping

Model mapping allows you to map request model names to actual models used by providers, supporting OpenAI compatibility, model aliases, and provider preferences.

Overview

Model mapping allows you to:

Use OpenAI model names with different providers
Create model aliases for easier invocation
Route specific models to preferred providers or accounts

Adding a Mapping

Navigate to Model Management > Model Mapping
Click Add Mapping
Configure the mapping rule
Save changes

Mapping Configuration

Basic Mapping

Field	Description
Request Model	Model name in the API request
Actual Model	Model name sent to the provider
Preferred Provider	(Optional) Specific provider to use
Preferred Account	(Optional) Specific account to use

Example

Map gpt-4 to DeepSeek-V3.2:

Request Model	Actual Model	Provider
`gpt-4`	`DeepSeek-V3.2`	DeepSeek

# Request with gpt-4
curl http://localhost:8080/v1/chat/completions \
  -d '{"model": "gpt-4", "messages": [...]}'

# Actually uses DeepSeek-V3.2 on DeepSeek

Wildcard Mapping

Use * as a wildcard to match multiple models.

Examples

Pattern	Matches
`gpt-*`	gpt-4, gpt-3.5-turbo, gpt-4o
`gpt-4*`	gpt-4, gpt-4o, gpt-4-turbo
`*-turbo`	gpt-3.5-turbo, gpt-4-turbo

Wildcard mappings are evaluated in order. More specific patterns should come first.

Priority Rules

When multiple mappings could match:

Exact match takes priority over wildcards
Earlier rules take priority over later ones
Provider-specific rules take priority over global rules

Example Priority

Order	Request Model	Actual Model	Priority
1	`gpt-4`	`DeepSeek-V3.2`	High (exact)
2	`gpt-*`	`GLM-5`	Low (wildcard)

Provider Preference

Route specific models to preferred providers:

Request Model	Actual Model	Provider
`gpt-4`	`DeepSeek-V3.2`	DeepSeek
`gpt-4`	`GLM-5`	GLM

If the preferred provider is unavailable, the system will try other providers.

Account Preference

Route to specific accounts within a provider:

Request Model	Actual Model	Provider	Account
`gpt-4-premium`	`DeepSeek-V3.2`	DeepSeek	Premium Account

Use Cases

OpenAI Compatibility

Map OpenAI model names to use with other providers:

Request	Actual
`gpt-4`	`DeepSeek-V3.2`
`gpt-3.5-turbo`	`GLM-5`
`gpt-4-vision`	`Qwen3-VL-235B-A22B`

Model Aliases

Create aliases for easier reference:

Request	Actual
`fast`	`DeepSeek-V3.2`
`smart`	`Kimi-K2.5`
`code`	`Qwen3-Coder`

Testing

Route test requests to specific accounts:

Request	Actual	Account
`test-model`	`DeepSeek-V3.2`	Test Account