Built-in provider directory: 62 presets + unlimited custom

The full list of factory-preset BYOK providers (snapshot 2026-07-15). Each ships with protocol wiring and a default model table — paste a key, test connectivity, go. Anything not listed connects as a custom endpoint. Keys stay encrypted on-device; requests go direct to the provider.

Frontier labs

ProviderPreset modalities
OpenAIText · Vision · Image · Speech
AnthropicText · Vision
GeminiText · Vision · Image · Video
Grok (xAI)Text · Vision · Image · Video · Speech
MistralText
PerplexityText

Mainland China majors

ProviderPreset modalities
DeepSeekText
ZhiPu GLMText
Alibaba Model Studio (Qwen)Text · Vision · Image · Video · Speech · Embedding
Moonshot (Kimi)Text
Doubao (Volcengine)Text · Vision · Image · Video · Embedding
Tencent HunyuanText
MiniMaxText
StepFunText
BaichuanText
01.AI YiText
Xiaomi MiMoText · Vision · Speech
Baidu QianfanText
Tencent Cloud TIText

CN GPU & open-model clouds

ProviderPreset modalities
SiliconFlowText · Vision · Image · Speech · Embedding
ModelScopeText
Infini-AIText
PPIOText
QiniuText
LANYUN MaaSText
CephalonText
SophNetText
CTYun XirangText
PH8Text
AlayaNewText
LongCatText

Aggregators & relays

ProviderPreset modalities
OpenRouterText · Vision · Image
AiHubMixText
ocoolAIText
DMXAPIText
AIOnlyText
BurnCloudText
TokenFluxText
302.AIText
PoeText
CherryINText
New API (self-hosted)Text
Vercel AI GatewayText

Fast-inference clouds

ProviderPreset modalities
GroqText
Together.AIText
FireworksText
NVIDIA NIMText
HyperbolicText
CerebrasText

Enterprise clouds & platforms

ProviderPreset modalities
Azure OpenAIText · Vision
Google Vertex AIText · Vision · Image · Video · Embedding
AWS BedrockText
GitHub ModelsText
GitHub CopilotText
Hugging FaceText

Local & self-hosted

ProviderPreset modalities
OllamaText · Image
LM StudioText
GPUStackText
OpenVINO Model ServerText

Media & embedding specialists

ProviderPreset modalities
KlingImage · Video
Jina AIEmbedding · Rerank
VoyageAIEmbedding · Rerank

Badges are preset declarations; actually available models come from each provider's live catalog fetched in-app. For wiring behavior and key policy, see the BYOK page and provider policy.

FAQ

My provider isn't listed — can I still use it?

Yes, via Add Custom Provider: any OpenAI- or Anthropic-compatible endpoint works with an API host and key (include the full version prefix, such as /v1). The 62 entries are factory presets; custom endpoints are unlimited.

What do the modality badges mean?

They are the capability flags declared by each preset template — text/vision/image/video/speech/embedding. The models actually available come from the provider's live model list fetched in-app; catalogs move fast, and the badge only means the protocol wiring is ready.

Where do my API keys live?

Encrypted on your device only. Requests go straight to the provider's official endpoint, never through Foreverse servers. Every provider row has a connectivity test button to verify key and network in one tap.

How do I connect local models (Ollama / LM Studio)?

Both are presets: start the server on your computer, point the app's API host at your LAN address (like http://192.168.x.x:11434/v1), keep phone and computer on the same Wi-Fi, and chat with zero API cost.