Remote Model Providers for Voxt
Voxt is local-first, not local-only. When you want a provider model, stronger remote rewrite, or an existing local server, you can configure a remote or OpenAI-compatible channel with the provider and credentials you choose.
The short version
- Remote providers are optional; they are not required for every local-first workflow.
- The documented catalog includes OpenAI, Anthropic, Google, DeepSeek, OpenRouter, xAI, Z.ai, Volcengine, Kimi, MiniMax, Alibaba Cloud, and others.
- Ollama, oMLX, and LM Studio are useful when the model runs in a separate local service but Voxt connects through an API.
- Provider availability, model IDs, billing, retention, and privacy policies belong to the provider and can change independently of Voxt.
Model map
| Model or family | Role | Useful when | Upstream source |
|---|---|---|---|
| OpenAI, Anthropic, Google, and DeepSeek Common provider channels for remote transcription, rewrite, conversation, and translation workflows. | Remote LLM or ASR | Use them when you prefer a hosted model and are comfortable with the provider's policy and billing model. | Voxt remote model documentation |
| OpenRouter, xAI, Z.ai, Volcengine, Kimi, and MiniMax Additional provider paths documented for model choice, regional availability, or specific model families. | Remote LLM | Choose them when a particular provider, model family, or endpoint fits your workflow. | Voxt remote model documentation |
| Ollama, oMLX, and LM Studio Separately hosted local model services that can expose an API for Voxt to call. | Local OpenAI-compatible service | Use them when you already manage local models outside Voxt or want a dedicated model server. | Ollama API documentation |
| Aliyun Bailian and StepFun ASR Remote speech recognition paths listed in the current Voxt provider documentation alongside OpenAI, Doubao, and GLM ASR. | Remote ASR | Compare them when language, endpoint, regional access, or provider policy is the deciding factor. | Voxt remote model documentation |
BYOK changes the data boundary
Bring-your-own-key means Voxt can call the provider you configure, but it does not make that provider local or change the provider's retention and training policies. Read the provider terms, use the privacy data-flow page as a map, and choose local channels when the content should not leave the Mac.
Remote provider versus local API service
Ollama, oMLX, and LM Studio can be local to your machine while still looking like a remote endpoint from the app's configuration layer. Keep that distinction explicit in setup guides: the network path is local, but the model lifecycle and server configuration live outside Voxt.
FAQ
Does Voxt require a remote AI provider?
No. Voxt supports local model channels and Apple System Dictation. Remote providers and BYOK credentials are optional choices for users who want a hosted model, a different model family, or a separately managed API service.
Can I use Ollama with Voxt?
The Voxt remote-model documentation includes Ollama as a local or OpenAI-compatible provider path. Configure the local server and model according to the current Voxt help guide and Ollama API behavior, then verify the selected channel with a non-sensitive test prompt.
Are remote model providers private?
Privacy depends on the provider, endpoint, model, account, and request content. Voxt does not turn a remote provider into a local workflow; review the provider policy and use Voxt's privacy data-flow documentation before sending sensitive audio or text.