LangChain Model Invocation -- init_chat_model() Function

LangChain init_chat_model() is one of the most commonly used functions in LangChain. It lets you connect to over 20 model providers in a unified way, without needing to remember each provider's class names and parameter differences.

Syntax

The syntax of the init_chat_model() function is as follows:

from langchain.chat_models import init_chat_model

# 完整语法
model = init_chat_model(
    model,                    # str | None：模型名称（provider:model 格式）
    *,
    model_provider=None,      # str | None：单独的模型提供商
    configurable_fields=None, # None | "any" | list[str]：可运行时修改的字段
    config_prefix=None,       # str | None：配置键前缀
    **kwargs,                 # 模型特定参数（temperature、max_tokens 等）
)

Parameter Description:

Detailed Explanation of the model Parameter

provider:model format (recommended)

The format isprovider:model name, separated by a colon:

Automatically Inferring the Model Provider

If you don't specify the provider prefix, LangChain will try to infer it from the model name:

Automatic inference is convenient, but not guaranteed to be 100% correct. For example, the gemini prefix may point to google_vertexai or google_genai, and future versions may change the default inference result. It is recommended to always use the provider:model format in production environments.

model_provider Parameter

When model_provider is specified separately, the effect is equivalent to the provider:model format:

Scenarios for using model_provider:

• When dynamically reading the provider name from a configuration file
• When the provider name and model name need to be configured independently (switched separately at runtime)
• When the model name cannot be automatically inferred and it is inconvenient to concatenate strings

Fixed Model vs Configurable Model

init_chat_model() has two usage modes:

Mode 1: Fixed Model

Specify a specific model string, returns a BaseChatModel instance that can be used directly:

If .env is in the current directory, use the following code to load the configuration from the current path:

import os

from dotenv import load_dotenv

# 加载当前目录 .env 文件
load_dotenv()
from langchain.chat_models import init_chat_model

# 指定了 model，返回固定模型
model = init_chat_model("deepseek:deepseek-v4-flash", temperature=0.7)
response = model.invoke("介绍Example")
print(response.content)

Mode 2: Configurable Model

Don't specify model (or set it to None) to create a model that can be dynamically switched at runtime:

Configurable models are very useful in A/B testing and cost optimization. You can switch models or adjust parameters by modifying the configuration without restarting the service.

All Supported Model Providers

The following are the providers natively supported by init_chat_model() and their installation packages:

Common kwargs Parameters

kwargs parameters are passed directly to the underlying model class. Commonly used ones include:

temperature and top_p are usually not set at the same time. temperature controls the "shape of the distribution", while top_p controls the "candidate range". For most scenarios, adjusting only temperature is sufficient.

ConfigurableModel—Runtime Model Switching

ConfigurableModel is an advanced usage of init_chat_model(), allowing you to dynamically specify models and parameters at runtime:

Possible values of configurable_fields:

When using configurable_fields="any", be careful about security: if you trust an insecure configuration source, sensitive fields such as api_key and base_url may be tampered with. In production environments, it is recommended to explicitly list the configurable fields.