Skip to main content
GPTfy - Salesforce Native AI Platform

BYOM (Bring Your Own Model)

An architecture letting enterprises plug their preferred LLM (Claude, GPT-4, Gemini, Llama) into Salesforce instead of being locked to the vendor's default.

Quick answer

What is BYOM (Bring Your Own Model)?

BYOM, or Bring Your Own Model, is an architecture that lets a company connect its own large language model to a platform instead of using the vendor's built-in one. In Salesforce, GPTfy routes prompts to Claude, GPT, Gemini, Llama or a self-hosted model through Named Credentials, so model choice, cost and hosting stay with the customer.

Last updated:

BYOM solves a critical enterprise concern: model lock-in. When a platform forces you to use only its native LLM, you're stuck with that vendor's pricing, performance, and security posture. BYOM architectures decouple the platform from the model, letting you choose the best LLM for each workload, Claude for long-context analysis, GPT-4 for general reasoning, Llama for on-prem deployments.

For Salesforce teams, BYOM matters because capability, price and hosting location vary widely between providers, and those variables shift every few months. A BYOM layer keeps that choice reversible. GPTfy implements it as an abstraction layer: prompts are formatted for the chosen model, PII masking is applied consistently, and responses are normalised before they return to Salesforce.

BYOM is especially important for: regulated industries needing on-prem models, cost-sensitive workloads benefiting from cheaper open-source LLMs, and teams wanting to A/B test models against the same prompts.

Frequently asked

BYOM (Bring Your Own Model), common questions

  • What does BYOM stand for?
    BYOM stands for Bring Your Own Model. It describes an architecture where a company connects a large language model it has chosen and licensed itself, rather than consuming whatever model the platform ships by default. The organisation keeps the provider contract, the API key and the decision about where inference runs.
  • How does BYOM work in Salesforce?
    A BYOM layer connects to the model provider through a Named Credential, which stores the endpoint and its authentication so no API key is written into Apex or a Flow. GPTfy formats the prompt, applies PII masking to the record data, makes the callout, and returns the response to the object, Flow or component that requested it.
  • Which models can you use with BYOM?
    GPTfy supports OpenAI, Anthropic Claude, Google Gemini, Meta Llama, DeepSeek, Perplexity and Grok, along with self-hosted deployments of open-weight models. Teams often route workloads differently: a long-context model for document analysis, a smaller cheaper model for classification, and a self-hosted model where data residency rules require inference to stay in a particular region.
  • Is BYOM more secure than a built-in model?
    Security depends on how it is configured, not on BYOM by itself. What BYOM changes is who holds the controls. Because the model contract is yours, you set data retention terms with your provider, choose the hosting region, and can require that raw data stays in Salesforce while only masked data reaches the model. Those levers are not available when the model is bundled.
  • What does BYOM cost compared with bundled AI?
    With BYOM you pay two separate bills: the platform licence and your model provider's usage. GPTfy charges a fixed $20/user/month, so platform cost stays predictable as usage grows, while inference is billed by your provider at their rates. Consumption-based bundles fold both into a single usage charge, which makes heavy usage harder to forecast.

Go deeper

Learn more about BYOM (Bring Your Own Model)

See GPTfy pricing