Use BigModel.io with LangChain

LangChain talks to BigModel.io through its OpenAI or Anthropic integration. Use ChatOpenAI for any key; use ChatAnthropic only if the key's group serves /v1/messages.

Model IDs come from the model catalog; availability depends on the key's group.

export BIGMODEL_API_KEY="sk-..."

#Install

pip install langchain-openai langchain-anthropic

npm install @langchain/openai @langchain/anthropic @langchain/core zod

#ChatOpenAI

import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="glm-5.3",
    base_url="https://bigmodel.io/v1",
    api_key=os.environ["BIGMODEL_API_KEY"],
    stream_usage=True,
)
print(llm.invoke("List five things to verify before launch.").content)

import { ChatOpenAI } from "@langchain/openai";

const llm = new ChatOpenAI({
  model: "glm-5.3",
  apiKey: process.env.BIGMODEL_API_KEY,
  configuration: { baseURL: "https://bigmodel.io/v1" },
  streamUsage: true,
});
const result = await llm.invoke("List five things to verify before launch.");
console.log(result.content);

In Python, stream_usage is off by default when you set a custom base URL, so streamed responses carry no usage_metadata. Pass stream_usage=True as shown.

Passing reasoning=... or use_responses_api=True switches ChatOpenAI to /v1/responses, which GLM and DeepSeek groups do not serve (404). Leave both unset for those keys.

#ChatAnthropic

import os
from langchain_anthropic import ChatAnthropic

llm = ChatAnthropic(
    model="glm-5.3",
    anthropic_api_url="https://bigmodel.io",
    api_key=os.environ["BIGMODEL_API_KEY"],
    max_tokens=1024,
)
print(llm.invoke("List five things to verify before launch.").content)

import { ChatAnthropic } from "@langchain/anthropic";

const llm = new ChatAnthropic({
  model: "glm-5.3",
  anthropicApiUrl: "https://bigmodel.io",
  apiKey: process.env.BIGMODEL_API_KEY,
  maxTokens: 1024,
});
const result = await llm.invoke("List five things to verify before launch.");
console.log(result.content);

ChatAnthropic also reads ANTHROPIC_API_URL / ANTHROPIC_BASE_URL from the environment; an explicit argument wins.

#Streaming

for chunk in llm.stream("Write a haiku about latency."):
    print(chunk.content, end="", flush=True)

const stream = await llm.stream("Write a haiku about latency.");
for await (const chunk of stream) {
  process.stdout.write(typeof chunk.content === "string" ? chunk.content : "");
}

#Tool calling

bind_tools sends standard OpenAI or Anthropic tool definitions; the gateway forwards them. Tool support depends on the model.

from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    """Get the current weather for a city."""
    return f"Sunny in {city}"

llm_with_tools = llm.bind_tools([get_weather])
msg = llm_with_tools.invoke("What's the weather in Hangzhou?")
print(msg.tool_calls)

import { tool } from "@langchain/core/tools";
import { z } from "zod";

const getWeather = tool(async ({ city }) => `Sunny in ${city}`, {
  name: "get_weather",
  description: "Get the current weather for a city.",
  schema: z.object({ city: z.string() }),
});

const msg = await llm.bindTools([getWeather]).invoke("What's the weather in Hangzhou?");
console.log(msg.tool_calls);

#reasoning_content caveat

Thinking models on GLM and DeepSeek groups return their reasoning in a reasoning_content field. The gateway passes it through, but ChatOpenAI does not surface it: you get the final answer only. If you need the reasoning text, call the model with the OpenAI SDK and read reasoning_content there. See Reasoning & thinking.

#Verify the setup

msg = llm.invoke("ping")
print(msg.content, msg.usage_metadata)

const msg = await llm.invoke("ping");
console.log(msg.content, msg.usage_metadata);

A reply with token counts means the base URL and key work. The request should also appear in Usage.

#Troubleshooting

404 on every request with ChatOpenAI. The base URL is missing /v1. In JS it must be inside configuration, not a top-level baseURL.

404 on every request with ChatAnthropic. The URL ends in /v1. Use https://bigmodel.io.

404 ... only supports /v1/chat/completions. Either ChatAnthropic is used with a Chat-Completions-only group, or ChatOpenAI was switched to the Responses API by reasoning / use_responses_api. Use plain ChatOpenAI.

No token usage on streamed responses. Set stream_usage=True (Python) or streamUsage: true (JS).

401 INVALID_API_KEY. The key is not reaching the client; an OPENAI_API_KEY or ANTHROPIC_API_KEY in your environment may be used instead. Pass api_key explicitly.

503 No available accounts: no available accounts supporting model: .... The model ID is wrong or not in the key's group. Copy it from the model catalog.

403 INSUFFICIENT_BALANCE or 429. Top up in Billing; for 429 see Limits & rate limits.

#Related

• OpenAI SDK and Anthropic SDK
• LlamaIndex and Vercel AI SDK
• Tool calling and Streaming