Skip to main contentSkip to content
    Log inBook a Working Session
    MODEL PROVIDERS · 15 INFERENCE · 12 EMBEDDING

    A customized model for each team, plus the providers you already approve.

    Each coworker on each dedicated team can get a model trained to that team's use case. You can still connect native APIs, cloud platforms or customer-controlled endpoints. Data never leaves your VPC. The same workspace and governance layer stay in place.

    See team models Explore inference providers
    NOT A GENERIC MODEL

    Each coworker, on each dedicated team, gets a customized model.

    Trained to your use case and how that team works. Data never leaves your VPC. Supported inference and embedding providers remain available below.

    01Per team

    A customized model for each dedicated team.

    Each coworker on each team gets a model trained to that team's use case and how they work, not a generic company chatbot.

    Per coworker · per team · your use case
    02Speed and cost

    More than 20x faster, with 70% lower token cost.

    Specialized team models respond more than 20x faster, with a 70% reduction in token cost and usage. High ROI on each token spent.

    20x faster · 70% token cost reduction
    03Your VPC

    Data never leaves your VPC.

    Payments does not train Platform. Platform does not train Support. Training and inference stay inside the tenant boundary.

    No cross-team leakage · tenant VPC
    ONE MODEL LAYER, THREE ROUTES

    Fit model access to your operating environment.

    Provider choice is configured at the model layer. Your enterprise connectors, retrieval scope and approval policies remain in place. A customized team model can run alongside these routes.

    01Native APIs

    Connect directly to leading model providers.

    Use provider-specific integrations for streaming, tool use and supported multimodal workloads.

    OpenAI · Anthropic · Gemini · Mistral · more
    02Cloud platforms

    Keep model traffic inside your cloud strategy.

    Route through Azure AI, Azure OpenAI or AWS Bedrock with deployment-specific credentials and endpoints.

    Azure AI · Azure OpenAI · AWS Bedrock
    03Controlled endpoints

    Run models on infrastructure you control.

    Connect Ollama or an OpenAI-compatible endpoint after validating reachability and model capabilities.

    Ollama · vLLM · LM Studio · compatible APIs
    INFERENCE PROVIDERS

    15 routes for generation, reasoning and tool use.

    Choose the provider and model that match the workload, deployment boundary and commercial requirements.

    OpenAI

    Supported provider

    GPT models through the native OpenAI API.

    Native APIStreamingTool use

    Anthropic

    Supported provider

    Claude models through the native Anthropic API.

    Native APILong contextTool use

    Google Gemini

    Supported provider

    Google multimodal models through the Gemini API.

    Native APIMultimodalTool use

    xAI

    Supported provider

    Grok models through the native xAI endpoint.

    Native APIStreamingTool use

    Mistral AI

    Supported provider

    Mistral-hosted models through its native API.

    Native APIStreamingTool use

    Cohere

    Supported provider

    Command models designed for enterprise AI workloads.

    Native APIEnterpriseTool use

    Groq

    Supported provider

    Low-latency inference for supported open models.

    Native APILow latencyTool use

    Fireworks AI

    Supported provider

    Hosted inference for open and custom model families.

    Native APIOpen modelsTool use

    Azure AI

    Supported provider

    Models deployed through Azure AI Foundry endpoints.

    Cloud platformPrivate endpointTool use

    Azure OpenAI

    Supported provider

    OpenAI model deployments hosted in Microsoft Azure.

    Cloud platformDeployment basedTool use

    AWS Bedrock

    Supported provider

    Multiple model families accessed through AWS.

    Cloud platformAWS credentialsMultimodel

    Ollama

    Supported provider

    Locally served models on customer-controlled infrastructure.

    Self-hostedLocal endpointPrivate

    Together AI

    Supported provider

    Hosted inference for open model families.

    Hosted modelsOpen modelsStreaming

    DeepSeek

    Supported provider

    DeepSeek reasoning and coding models through its API.

    Compatible APIReasoningCoding

    OpenAI Compatible

    Supported provider

    Custom endpoints such as vLLM, LM Studio and compatible gateways.

    Custom endpointSelf-hostedFlexible
    EMBEDDING PROVIDERS

    12 routes for retrieval and semantic search.

    Keep retrieval flexible with managed APIs, cloud-hosted deployments, specialized retrieval models or local endpoints.

    OpenAI

    Supported provider

    OpenAI text embedding models.

    Managed APIConfigurable dimensions

    Google Gemini

    Supported provider

    Google text embedding models.

    Managed APIGoogle ecosystem

    Cohere

    Supported provider

    Cohere embeddings for search and multilingual retrieval.

    RetrievalMultilingual

    Mistral AI

    Supported provider

    Mistral embedding models through its native API.

    Managed APINative integration

    Azure AI

    Supported provider

    Embedding deployments served through Azure AI Foundry.

    Cloud platformCustom endpoint

    Azure OpenAI

    Supported provider

    OpenAI embedding deployments hosted in Azure.

    Cloud platformDeployment based

    AWS Bedrock

    Supported provider

    Embedding model families accessed through AWS.

    Cloud platformAWS credentials

    Ollama

    Supported provider

    Local embedding models on customer-controlled infrastructure.

    Self-hostedLocal endpoint

    Together AI

    Supported provider

    Hosted embedding models for retrieval workloads.

    Hosted modelsRetrieval

    Voyage AI

    Supported provider

    Retrieval and code-focused embedding models.

    RetrievalCode search

    Jina AI

    Supported provider

    Multilingual embeddings and late-interaction retrieval models.

    MultilingualRetrieval

    OpenAI Compatible

    Supported provider

    Custom embedding endpoints that implement the compatible API.

    Custom endpointSelf-hosted
    CONTROL THE ROUTE

    Model flexibility without losing operational control.

    Linkence keeps provider configuration explicit, testable and independent from the systems that supply company context.

    01

    Bring approved credentials

    Use customer-managed provider credentials where the deployment supports them.

    02

    Set each model layer

    Choose inference and embedding routes independently for the workload.

    03

    Test before publishing

    Validate connectivity and model configuration before the workspace uses the route.

    04

    Keep connectors unchanged

    Change model configuration without rebuilding the company systems connected to Linkence.

    MODEL PROVIDER QUESTIONS

    Direct answers about model choice.

    Availability and operating boundaries are confirmed for the selected deployment.

    Does Linkence train a customized model for each team?+

    Yes. Each coworker on each dedicated team can get a customized model trained to that team's use case and how they work. Payments does not train Platform. You can still use supported inference and embedding providers, including native APIs, cloud platforms and customer-controlled endpoints. Data never leaves your VPC. Specialized team models respond more than 20x faster, with a 70% reduction in token cost and usage.

    Is Linkence model-agnostic?+

    Yes. Linkence separates the governed workspace from the configured model route. Administrators can choose from supported inference and embedding providers without reconnecting company systems.

    Can inference and embeddings use different providers?+

    Yes. The inference and embedding layers are configured independently. A workspace can use one provider for response generation and another for retrieval embeddings.

    Can we use our own model credentials?+

    Customer-managed provider credentials are supported in approved configurations. Exact key ownership, network routing and operational responsibilities depend on the selected deployment.

    Can Linkence connect to self-hosted models?+

    Yes. Ollama and OpenAI-compatible endpoints support customer-controlled model serving. Endpoint reachability, model capabilities and production support are verified during configuration.

    Are all providers available in every deployment?+

    Provider availability can vary by deployment, region and commercial configuration. Linkence confirms the supported route before production access is granted.

    DESIGN THE RIGHT MODEL ROUTE

    Choose providers around one real workload.

    We will map the inference, embedding, credential and deployment requirements for your workspace.

    Book a technical review View all integrations
    Provider availability is confirmed before production access.