Models and costs

Which model to choose, and what it costs

Libris Translate does not include a model: you bring one and pay its provider directly. Here is what a real book cost, and how to get started in five minutes.

Prices read on 30 September 2026 on each provider’s page, in US dollars, before tax. They change: Libris Translate uses the prices you enter for each provider.

What a real book cost

Measured on 1 October 2026, Libris Translate 0.25.0, quality “high”, default autopilot: Alice’s Adventures in Wonderland (public domain), English → French, 26,530 words, translated end to end. 2,695 model calls, 39,832,243 tokens sent and 2,026,109 received, analysis, checks, three autopilot rounds, escalations and failed calls included. Translation model: GPT-6 Luna; GPT-6 Sol for the passages the autopilot escalates (wordplay, arbitration). The book was translated with a ChatGPT subscription through the Codex bridge: the amounts below apply the APIs’ list prices to the same tokens.

Model cost for 100 000 words, at list prices
ProviderModelInput / output, per million tokensFor
As measured (Luna, and Sol for escalations) $107
OpenAI GPT-6 Luna $0.10 / $0.50 * $19
OpenAI GPT-6.1 Sol $2.00 / $10 * $377
Anthropic Claude Haiku 4.5 $1.00 / $5.00 * $188
Anthropic Claude Sonnet 5.5 $2.00 / $10 * $377
Google Gemini 3.8 Flash $0.75 / $3.75 * $141
Google Gemini 3.5 Flash-Lite $0.30 / $2.50 * $64
Mistral Mistral Small 4 $0.15 / $0.60 * $27
Mistral Mistral Large 3 $0.50 / $1.50 * $87
DeepSeek DeepSeek-Flash $0.30 / $1.20 $54

* Half price in batch, for books that can wait.

Every model counts tokens its own way: the same text can take up to about 30 % more tokens on another model, and a model that writes longer answers costs more. Treat the other rows as an order of magnitude, and let Libris Translate estimate your own book before starting. Estimate before starting · Batch processing at half price

The model is the only cost that grows with the book. The Libris Translate licence is separate; the free trial covers 100,000 words a month for 30 days.

Which model to choose

  • To start: an inexpensive, recent model such as GPT-6 Luna, the one we measure below. Most books do not need more. What the AI book translator does with it
  • For the hard passages: a higher model (GPT-6 Sol, Claude Sonnet, Gemini Pro) as the judge or for escalations only, not for the whole book. Autopilot
  • For your own hardware: any server that speaks the OpenAI-compatible Chat Completions protocol (vLLM, llama.cpp, Ollama…); the cost is its electricity and hardware.
  • With a ChatGPT subscription: the optional Codex bridge translates within the subscription’s usage limits instead of paying per token.
  • Before choosing: have two or three models translate the same passages of your book, side by side. Compare providers

Get an API key in five minutes

Then, in Libris Translate: Settings › LLM providers › Add, choose the connection below, paste the key, and click “Test / detect models”. LLM providers

  • OpenAI

    1. Create an account on platform.openai.com.
    2. Add prepaid credit in Billing (a few dollars are enough to try).
    3. Create a key in API keys and copy it at once.

    In Libris: OpenAI Chat Completions or Responses with an API key.

    OpenAI — quickstart

  • Anthropic

    1. Sign in on platform.claude.com.
    2. Add a payment card in Billing.
    3. Settings › API keys › Create key; the key (sk-ant-…) is shown once.

    In Libris: Anthropic Claude with an API key.

    Anthropic — get an API key

  • Google Gemini

    1. Open aistudio.google.com/apikey with a Google account.
    2. Accept the terms, then Create API key.
    3. Turn on billing for the paid tier: on the free tier, Google may use your texts to improve its products.

    In Libris: OpenAI-compatible Chat Completions, address https://generativelanguage.googleapis.com/v1beta/openai/.

    Google — Gemini API key

  • Mistral

    1. Create an account on console.mistral.ai.
    2. Choose a plan in the console (a free tier exists).
    3. API keys › Create new key.

    In Libris: OpenAI-compatible Chat Completions, address https://api.mistral.ai/v1.

    Mistral — quickstart

  • DeepSeek

    1. Create an account on platform.deepseek.com.
    2. Top up the balance.
    3. API keys › Create new API key.

    In Libris: OpenAI-compatible Chat Completions, address https://api.deepseek.com.

    DeepSeek — API docs

  • OpenRouter

    1. Create an account on openrouter.ai.
    2. Buy credits (the card fee is 5.5 %).
    3. Settings › Keys › Create key; one key reaches many providers.

    In Libris: OpenAI-compatible Chat Completions, address https://openrouter.ai/api/v1.

    OpenRouter — FAQ