AI Translation & MT Pricing

The choice depends on whether you need fixed speed or creative control. Traditional MT (like Systran) is faster and more cost-effective for high-volume, straightforward text, but it cannot be “prompted” because it simply translates the input as provided. LLMs (like GPT-5 or Gemini) are better for context-heavy content, creative marketing, or when you need the model to follow specific style instructions through custom prompts.

According to recent WMT 2025 research, models like Gemini 2.5 Pro now lead in overall quality across many language pairs, often outperforming both traditional NMT and other LLMs in professional and technical domains. However, because performance varies by language and domain, the simplest experience is to use our Best Engine feature, which automatically selects the top-performing model for your specific project.

Imagine you are on the Custom.MT Starter Edition (€200/month) and you translate 12 million characters in a month using DeepL via our Default Keys. Your bill would be calculated in three parts:

1. Subscription Base: €200 (covers the platform and your first 10M characters).
2. Capacity Overage: You processed 12M characters, which is 2M over your limit. At the Starter overage rate (€20 per 1M), this adds €40.
3. Translation Usage (DeepL): You pay the provider rate for all 12 million characters. At the DeepL rate of €0.000030 per character, this adds €360.

Total Monthly Bill: €600 (€200 + €40 + €360).

Note: If you used your own “User Keys” for DeepL, your Custom.MT bill would only be €240, and you would pay DeepL separately for the translation volume.

Custom.MT separates platform costs from provider costs. One invoice covers your subscription; the second reflects the actual usage of the MT and LLM providers.

The Custom.MT Console includes a Statistics section where you can track your daily character and token consumption.

Yes. Providers charge for every bit of data sent to their servers. This includes the source text, any custom prompts you create in our Prompt Studio, and glossary terms used to ensure accuracy. Because these elements require computational power from the AI/MT engine, they contribute to the total character or token count.

To ensure your translated files open correctly in CAT tools or web browsers, we must send the HTML/XML tags (like <b> or <div>) along with the text. Providers charge for these characters because they must be processed to maintain the document’s structure. Excluding them would lower the bill but break your file formatting.

No. The monthly processing capacity included in your plan (e.g., 10M characters for Starter) resets at the beginning of each billing cycle. If you don’t use the full amount, the remaining balance does not carry over. If you exceed it, you simply pay the flat overage rate defined in your plan.

The Starter Edition is designed for small translation teams and LSPs who need essential AI tools and glossary support for moderate volumes. The Advanced Edition fits growing companies localizing over 10 million words per year that require custom prompts and API access for integrations. The Compliance Edition is built for large enterprises with strict data privacy requirements that need features like PII tokenization, dedicated hosting, or a branded Corporate Translation Portal.

The main concern for most teams is outgrowing a plan too quickly or overpaying for capacity they don’t use. At Custom.MT, we recommend starting based on your security and feature requirements rather than just volume. Since every plan allows for uninterrupted processing even if you exceed your limit, you are never “locked out” of your work. If you need basic AI glossary support, start with Starter. If you need to build custom prompts or use our API, choose Advanced. If your priority is data privacy or a branded portal for your employees, Compliance is the correct fit. You can always monitor your actual usage in the Console and adjust your subscription as your localization needs change.