> ## Documentation Index > Fetch the complete documentation index at: https://docs.ai-stats.phaseo.app/llms.txt > Use this file to discover all available pages before exploring further. # Exploring API Providers > Understand what API providers do and how to evaluate them for reliability, performance, and cost. API providers are the platforms that host and serve model APIs. Model organizations train models. Providers expose those models through production endpoints, authentication, usage controls, and operational tooling. ## What an API provider typically handles * Authentication and key management * Request routing and availability management * Usage metering and billing surfaces * API and SDK ergonomics * Regional hosting and compliance controls ## Provider vs organization | Aspect | Organization | API Provider | | ------------ | -------------------------- | --------------------------------------------- | | Primary role | Builds and trains models | Hosts and serves model APIs | | Focus | Research and model quality | Reliability, access, and operations | | Distribution | Usually own model family | May expose models from multiple organizations | Many companies are both, depending on the product surface. ## What you can inspect in AI Stats provider pages | Section | What it tells you | | -------- | --------------------------------------------------------- | | Overview | Provider identity, docs links, and supported capabilities | | Models | Which models and variants are available | | Pricing | Provider pricing normalization and meters | | Limits | Request and throughput limits when available | | Features | Streaming, tools, structured output, and modality support | ## AI Stats Gateway as a provider layer AI Stats Gateway acts as a unified provider layer across many backends. Benefits: * Unified request and response shape across providers * Provider-aware routing and fallback behavior * Unified SDK usage for multi-provider integrations * Centralized usage and cost observability * Optional BYOK support for provider-native keys ## Choosing a provider When evaluating providers, focus on: * Model availability for your use case * Reliability and latency in your target region * Cost profile at your expected traffic level * Feature parity for tools, streaming, and structured outputs * Compliance and data handling requirements ## Example evaluation questions * Which providers expose the model family I need? * Which provider gives better latency for my region? * What is the lowest-cost option at my token volume? * Which provider supports my required endpoint features? ## Next steps Compare costs, speeds, and efficiency metrics across providers.