Models

Every chat completion request names a model in its model parameter. The List models endpoint returns the models you can use, and Retrieve a model returns one of them.

Model ID	Created	Knowledge cutoff	Context window	Max output
`Beam-501B-A23B`	October 5, 2026	June 30, 2026	256K¹	128K

¹ The context window may change during the beta.

Beam supports reasoning, tool calling, and structured outputs.

List models

curl https://api.reflection.ai/openai/v1/models \
  -H "Authorization: Bearer $REFLECTION_API_KEY"

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.reflection.ai/openai/v1",
    api_key=os.environ["REFLECTION_API_KEY"],
)

for model in client.models.list():
    print(model.id)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.reflection.ai/openai/v1",
  apiKey: process.env.REFLECTION_API_KEY,
});

for await (const model of client.models.list()) {
  console.log(model.id);
}

Response

{
  "object": "list",
  "data": [
    {
      "id": "Beam-501B-A23B",
      "object": "model",
      "created": 1791158400,
      "owned_by": "reflection",
      "shutdown_date": null,
      "context_length": 262144,
      "max_output_tokens": 131072,
      "reasoning": {
        "supported_efforts": ["max", "xhigh", "high", "medium", "low"],
        "default_effort": "medium",
        "mandatory": true
      }
    }
  ]
}

Each model object has these fields:

Field	Description
`id`	The identifier to pass as `model` in a request.
`object`	Always `model`.
`created`	The Unix time, in seconds, when the model was published, or `0` if its publication time isn’t announced.
`owned_by`	The organization that owns the model.
`shutdown_date`	The date the model will shut down, or `null` if none has been announced.
`context_length`	The largest request the model supports, in tokens, counting the prompt and the generated output together. A request longer than the model accepts is rejected, not truncated. Omitted if not announced.
`max_output_tokens`	The most tokens the model supports generating in response to one request. Omitted if not announced.
`reasoning`	The model’s reasoning efforts: the `reasoning_effort` values it accepts, the effort it applies when a request omits one, and whether reasoning is always on. Omitted if not announced.

Retrieve a model

To get one model, request it by ID. The response is a single model object with the same fields, and an unknown or unavailable model returns 404 with code model_not_found.

curl https://api.reflection.ai/openai/v1/models/Beam-501B-A23B \
  -H "Authorization: Bearer $REFLECTION_API_KEY"

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.reflection.ai/openai/v1",
    api_key=os.environ["REFLECTION_API_KEY"],
)

model = client.models.retrieve("Beam-501B-A23B")
print(model.id, model.owned_by)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.reflection.ai/openai/v1",
  apiKey: process.env.REFLECTION_API_KEY,
});

const model = await client.models.retrieve("Beam-501B-A23B");
console.log(model.id, model.owned_by);

Model lifecycle

When a model is scheduled to shut down, its shutdown_date is set. Plan to move to another model before that date; requests that name a model that no longer exists return 404. To detect upcoming shutdowns automatically, check shutdown_date on the models you use, for example at application startup or in a scheduled job.

Get started

Coding agents

Mirror CLI

Guides

Resources

List models

Retrieve a model

Model lifecycle

​List models

​Retrieve a model

​Model lifecycle

List models

Retrieve a model

Model lifecycle