Skip to content

Translation Models

Last updated: August 25, 2026

OmniTranslate offers several models with different trade-offs in speed, quality, and cost. The model selector in the app is the source of truth for which models are currently available to you.

The model list is provided by OmniTranslate rather than being fixed in the web or mobile app. This lets us add or retire models that use compatible pricing and translation capabilities without requiring an app update. If the selector looks out of date, update the app and try again.

Reading the Model Selector

Each model shows a compact price for comparing models before you translate:

Selector labelMeaning
3 /1K charsThe listed price is 3 credits per 1,000 characters.
~9 /1K charsA typical translation costs about 9 credits per 1,000 characters. The final price can vary.

The ~ marks a comparison estimate rather than a fixed price. Actual usage varies with the source language, glossary and context, generated translation, and model reasoning.

Current Models

OmniTranslate models

These models have a fixed listed price per 1,000 characters.

ModelBest forPrice
Wu 1.5 FlashFast translations with a strong quality-to-cost balance3 credits /1K chars
Wu 1.6 Core PreviewEveryday translation and the recommended default6 credits /1K chars
Wu 1.6 Nova PreviewHigher fidelity for difficult chapters20 credits /1K chars
Wu Ultra ExperimentalOur highest-quality experimental model60 credits /1K chars

Wu Ultra is a Premium model and uses credits rather than daily or weekly quota.

Gemini and DeepSeek models

These models are priced from the input and output tokens used by a completed translation. The selector shows a rounded /1K chars estimate so that they remain easy to compare with other models.

ModelTypical selector estimateInput priceOutput price
Gemini 3.5 Flash Lite~9 credits /1K chars900 credits /1M tokens7,500 credits /1M tokens
Gemini 3.7 Flash~18 credits /1K chars2,250 credits /1M tokens11,250 credits /1M tokens
Gemini 3.1 Pro~150 credits /1K chars6,000 credits /1M tokens36,000 credits /1M tokens
DeepSeek V4 Flash~9 credits /1K chars660 credits /1M tokens1,980 credits /1M tokens

Model names marked Preview or Experimental may change as we continue evaluating them.

How Token Pricing Is Calculated

For a token-priced model, the final translation price is:

text
(input tokens × input price + output tokens × output price) / 1,000,000

Input tokens cover the text, instructions, glossary, and context used for the translation. Output tokens include the translated text and any reasoning or thinking used by the model, rather than only the text you can see.

For example, a Gemini 3.7 Flash translation using 4,000 input tokens and 2,000 output tokens costs:

text
(4,000 × 2,250 + 2,000 × 11,250) / 1,000,000 = 31.5 credits

Token counts are not the same as character counts, which is why two chapters of the same length can have different final prices.

What You Pay

For a model without ~, the listed /1K chars price applies. For a model with ~, the final price is calculated from the input and output tokens used by your translation and may be higher or lower than the selector estimate.

See Credits & Usage for how quota, credits, and failed translations are handled.

Translation Add-ons

Advanced Term Extraction adds 2 credits per 1,000 characters on top of the selected model's translation price.

Which Model Should You Choose?

  • Start with Wu 1.6 Core for everyday translation.
  • Choose Wu 1.5 Flash when speed and lower cost matter most.
  • Try Wu 1.6 Nova for difficult chapters where higher fidelity matters.
  • Use Wu Ultra when you want the highest-quality OmniTranslate model and are comfortable with its experimental status.
  • Use a Gemini or DeepSeek model when you want an alternative, remembering that its final price can differ from the selector estimate.