Deference
Deference is an AI inference gateway and model marketplace that gives developers one API for accessing hundreds of language, vision, image, embedding, and other models. Its stated differentiator is usage in one place with no Deference markup, alongside privacy-oriented routing such as Incognito.
At a Glance
- Individual developers
- AI application and API developers
- Coding-agent users
- Teams and enterprise organizations using shared connectors and budgets
AI Tools by Deference
(1)
Deference Incognito
Confidential LLM Inference API Route
Discussions
No discussions yet
Be the first to start a discussion about Deference
Products & Services
An OpenAI-compatible API and dashboard for accessing a broad catalog of models through a single Deference key. The documentation shows support for chat completions, Responses-style integrations, Python and TypeScript SDK usage, and coding-tool configuration.
A privacy-focused route on the Deference API. Appending :incognito to a supported model ID sends the request only to Phala confidential-computing hardware (Intel TDX), with zero-data-retention routing and no fallback to another provider.
An MCP server that lets coding agents work with a Deference account; the agent can check balance/credit, select models, inspect pricing and spending, and run models within a user-set budget.
Market Position
Deference positions itself as a unified, no-markup model gateway with strong coding-agent support and a privacy differentiator. It overlaps with multi-model gateways such as OpenRouter, while differentiating through Incognito confidential-computing routes, budget-aware MCP tooling, and a single workflow for model discovery, usage, and spending.
Business Model
Revenue Model
Pay-per-request inference. Deference's terms and pricing documentation state that requests are charged at OpenRouter's model price plus OpenRouter's 5% platform fee; Deference adds no markup and there is no subscription requirement. Requests are paid from account credit.
Pricing Tiers
OpenRouter model price plus OpenRouter's 5% platform fee; no Deference markup and no subscription.
Target Markets
- Individual developers
- AI application and API developers
- Coding-agent users
- Teams and enterprise organizations using shared connectors and budgets
- Developers integrating multiple LLM and multimodal models through one API
- Coding agents and IDE-based AI development workflows
- Agentic workflows that need model choice, budget checks, and tool calling
- Privacy-sensitive inference using confidential computing
- Teams that want centralized model access, usage tracking, and spending controls