YandexGPT 5 Lite API: high-volume short tasks and pricing

Short answer: the model id is yandex/yandexgpt-5-lite, the request is OpenAI-shaped

LLMTR (llmtr.com) is a gateway that gives access to language models hosted in Turkey and at global providers through a single OpenAI-compatible API. You call YandexGPT 5 Lite there with the model id yandex/yandexgpt-5-lite on the /v1/chat/completions endpoint at https://llmtr.com/v1.

YandexGPT 5 Lite is the lightweight release of YandexGPT 5; the LLMTR catalog describes it for chat and user-facing work that needs fast, steady short answers. Its context window is 32,768 tokens; it takes text and returns text. Input and output bill at one rate, $1.639344 per 1M tokens. That makes cost easy to work out in advance for jobs made of many small requests: classification into a fixed label set, routing an incoming message to the right queue, and short replies.

• Yandex AI Studio, available generative models · 2026-10-12

How to use the YandexGPT 5 Lite API for high-volume short tasks

High-volume short tasks follow one pattern: a short, fixed system prompt, one user message and a one-word answer. The Python example below sorts a support message into one of five labels. The request uses temperature 0 and max_tokens 10; max_tokens limits the output to roughly label length.

The code makes two checks. If finish_reason is length, the answer was cut off and the label may be incomplete, so the message goes to the fallback path. Otherwise the returned text is lowercased and compared with the label list, and any answer outside the list also goes to the fallback path. The fallback path can be a human queue or a separate review step. That way an unexpected word from the model never sends a message to the wrong queue.

A support-message classifier on yandex/yandexgpt-5-lite with the OpenAI Python SDK; it checks finish_reason and the label list and returns only the label.

import os
from openai import OpenAI

client = OpenAI(base_url="https://llmtr.com/v1", api_key=os.environ["LLMTR_API_KEY"])

LABELS = ["billing", "delivery", "refund", "technical", "other"]
SYSTEM = "Classify the support message. Reply with exactly one label from: " + ", ".join(LABELS) + "."


def classify(message: str) -> str:
    response = client.chat.completions.create(
        model="yandex/yandexgpt-5-lite",
        messages=[
            {"role": "system", "content": SYSTEM},
            {"role": "user", "content": message},
        ],
        temperature=0,
        max_tokens=10,
    )
    choice = response.choices[0]
    if choice.finish_reason == "length":
        return "fallback"  # the answer was cut off
    label = (choice.message.content or "").strip().strip(".").lower()
    return label if label in LABELS else "fallback"


print(classify("I was charged twice for the same order."))

YandexGPT 5 Lite pricing: input and output at one rate

On LLMTR, yandex/yandexgpt-5-lite costs $1.639344 per 1M tokens for both input and output. This equals Yandex's published USD list price; Yandex quotes the price per 1,000 tokens and the catalog shows it per 1M tokens. The price was checked on 2026-10-12. No margin is added to this model's price; LLMTR's 8% platform fee is charged only when you top up credit.

Because input and output share a rate, cost is a single multiplication: total tokens x $1.639344 / 1M. A worked example: 100,000 support messages, each with 250 input and 10 output tokens, come to 26,000,000 tokens. That is 26,000,000 x $1.639344 / 1M = $42.622944. The system prompt is sent as input with every request, so count it inside those 250 tokens; this model has no cache discount.

• Yandex AI Studio, pricing · 2026-10-12

Practical tips for high volume

Keep the system prompt short and fixed. It is resent with every message and billed as input, so every 100 tokens you add to it mean 10,000,000 extra tokens, or $16.39344, across 100,000 messages.

Keep max_tokens small. A label is a few tokens, and a small max_tokens limits output cost when the model starts writing an explanation. Always validate the returned label against your own list and send unknown answers to the fallback path. If finish_reason often comes back as length, raise max_tokens slightly.

If you need one more short field next to the label, such as a priority, ask for JSON with response_format {"type": "json_object"} and state in the prompt that you expect JSON; Yandex's formatting page recommends the same. For short user-facing replies you can send stream: true; usage arrives at the end of the stream with stream_options.include_usage.

• Yandex AI Studio, formatting model responses · 2026-10-12

Limits: text only, no reasoning_effort

YandexGPT 5 Lite accepts text only. A request that contains an image is refused with 400 unsupported_input and is not charged. The reasoning_effort field is not supported either: a request that sends it gets 400 invalid_request, so make sure your client does not add the field by default.

A 32,768-token context leaves plenty of room for short messages; split long documents into parts before you send them. Tool calling works in the OpenAI tools format, so when a message should start an action you can define a tool instead of a label.

Getting started on LLMTR: account, key and first request

Step 1: create an account at llmtr.com and top up prepaid credit. Step 2: create a key under Dashboard > API Keys; the raw key is shown only once, so store it in an environment variable such as LLMTR_API_KEY. Step 3: in your OpenAI-compatible SDK, set base_url to https://llmtr.com/v1, use yandex/yandexgpt-5-lite as the model and send the first request.

Each request is deducted from your credit at the catalog price in effect when it is processed; the same key works for every model in the catalog.

Как использовать YandexGPT 5 Lite API для массовых коротких задач

LLMTR (llmtr.com) — совместимый с OpenAI API-шлюз. Модель yandex/yandexgpt-5-lite вызывается через /v1/chat/completions по адресу https://llmtr.com/v1: в OpenAI SDK достаточно указать base_url, ключ и идентификатор модели.

YandexGPT 5 Lite — облегчённая версия YandexGPT 5; в каталоге LLMTR она описана для чатов и пользовательских сценариев, где нужны быстрые и стабильные короткие ответы. Типичные массовые короткие задачи: классификация по фиксированному набору меток, маршрутизация обращений, короткие ответы. Контекст — 32 768 токенов. Вход и выход стоят одинаково: 1,639344 $ за 1 млн входных токенов и 1,639344 $ за 1 млн выходных, поэтому стоимость равна общему числу токенов, умноженному на эту ставку. Например, 100 000 сообщений по 250 входных и 10 выходных токенов обойдутся в 42,622944 $.

Модель принимает только текст: запрос с изображением отклоняется с ошибкой 400 unsupported_input и не тарифицируется. Параметр reasoning_effort не поддерживается (400 invalid_request). Если finish_reason равен length, ответ обрезан; если это происходит часто, немного увеличьте max_tokens.

Практический совет: держите системный промпт коротким и неизменным, задайте небольшой max_tokens, а полученную метку сверяйте со своим списком; неизвестные ответы отправляйте на резервную обработку.

Чтобы начать, выполните три шага:

• Создайте аккаунт на llmtr.com и пополните баланс.
• В разделе Dashboard > API Keys создайте ключ: он показывается один раз, поэтому сохраните его в переменной окружения LLMTR_API_KEY.
• Укажите base_url https://llmtr.com/v1, модель yandex/yandexgpt-5-lite и отправьте первый запрос.

• Yandex AI Studio, Доступные генеративные модели · 2026-10-12

Call YandexGPT 5 Lite through LLMTR

Create an account, create an API key and send the first request to https://llmtr.com/v1 with yandex/yandexgpt-5-lite.

1. Account. Create an account at llmtr.com and top up prepaid credit.
2. API key. Create a key under Dashboard > API Keys. The raw key is shown once; store it as the LLMTR_API_KEY environment variable.
3. First request. In an OpenAI-compatible SDK, set base_url to https://llmtr.com/v1 and send the first request with yandex/yandexgpt-5-lite as the model.

Frequently asked questions

How much does the YandexGPT 5 Lite API cost?

On LLMTR, input and output cost the same: $1.639344 per 1M tokens. The context window is 32,768 tokens.

How much does it cost to classify 100,000 short messages with YandexGPT 5 Lite?

With 250 input and 10 output tokens per message, the total is 26,000,000 tokens and the cost is $42.622944. Count the system prompt inside those 250 tokens.

Does YandexGPT 5 Lite accept images or reasoning_effort?

No. The model accepts text only; a request with an image is refused with 400 unsupported_input and is not charged. A request that sends reasoning_effort gets 400 invalid_request.

Related posts

• Yandex models compared: Alice AI LLM, Flash, YandexGPT Pro and Lite
• LLM API for Russian text: summarizing and classifying

• YandexGPT 5 Lite model
• Yandex documentation
• Chat Completions guide
• Streaming documentation