SubEnvoy’s server-side translation engine has changed again. Standard mode now uses DeepSeek V4 Flash 0731, while Advanced mode has moved from Gemini to Qwen 3.7 Plus.
This is a backend update, so there is no new App version to install. The new models are already used whenever you start a translation.
The Models Behind Each Mode
Standard mode now runs on DeepSeek V4 Flash 0731. It remains the practical choice for everyday subtitle translation, where speed, multilingual consistency, and cost efficiency all matter across hundreds or thousands of subtitle lines.
Advanced mode now runs on Qwen 3.7 Plus. It replaces Gemini 3.1 Flash Lite for translations that benefit from broader context and deeper reasoning, including dialogue with idiomatic language, recurring terminology, or meaning that spans several subtitle lines.
The purpose of the two modes has not changed. Standard mode is designed for efficient daily use, while Advanced mode spends more resources on difficult material.
Why We Could Not Continue Using Gemini
The reason for this change is operational rather than cosmetic.
OpenRouter recently began applying access restrictions to our account based on the region associated with its billing address. As a result, we can no longer use closed commercial models offered by US companies through that account, including models from OpenAI, Anthropic, and Google.
Advanced mode previously depended on Gemini. Once that access was no longer available, continuing with the same model was not an option.
We could have treated this as a temporary inconvenience and looked for a short-lived workaround. Instead, we chose to address the larger issue: a production service should not depend on model access that can disappear because of an external billing-region policy.
Building for Continuity
SubEnvoy is an independent product. Its translation service has to remain financially and operationally sustainable, not just technically impressive on the day a model is selected.
That means evaluating models by more than a familiar brand name. Translation quality remains essential, but so do reliable access, predictable pricing, throughput, multilingual performance, and the ability to keep serving users over the long term.
DeepSeek V4 Flash 0731 and Qwen 3.7 Plus are the current result of that evaluation. They allow both translation modes to continue without changing how you use SubEnvoy.
Our Direction From Here
Going forward, SubEnvoy will no longer build its translation service around access to closed commercial models from US providers. We will prioritize capable models from the broader open-model ecosystem and providers whose access terms make long-term operation possible.
This does not mean choosing a model simply because it is described as open, nor does it mean claiming that open models are automatically better. Every candidate still has to earn its place through translation quality, consistency, speed, availability, and cost.
The specific models may continue to evolve as the ecosystem changes. The commitment is not to a permanent model name; it is to a translation service that remains dependable, transparent, and sustainable.
No App update is required. Choose Standard or Advanced mode as usual, and SubEnvoy will handle the rest on the server.
Breaking language barriers, opening a vision without borders.