# Liquid Inference Liquid Inference routes each AI request through a provider auction, matching your model, region, speed, and quality requirements with a competitive offer. - Findings24: - Website: - Pricing: Freemium - Categories: AI Gateway And Routing, AI Inference Platform ## What is Liquid Inference? Liquid Inference is an auction exchange for large language model inference from Architect. It lets model providers quote prices for the models they serve, while buyers define requirements such as model selection, region, speed, and minimum quality. For each request, eligible providers compete and the selected offer is used to generate the response. The service is designed for teams sending AI requests through an API or compatible coding and agent tools. Its routing workflow exposes live offers, supports per-request spending caps, and charges for tokens used. Buyers can also review itemized request costs and public price history. ## What can Liquid Inference do? ### Per-request provider auction Providers post a price per token for the models they serve. Eligible offers compete for each request, and the lowest-priced offer that meets the buyer’s requirements is selected. ### Configurable routing terms Buyers can specify models, regions, speed, and minimum quality before sending requests, limiting the auction to providers that satisfy those terms. ### API-compatible integration Liquid Inference provides an API endpoint and documents a base-URL change for clients using the OpenAI or Anthropic API formats. ### Spending caps and usage billing Requests can use a price cap, while billing is based on tokens used. Itemized billing shows the cost of individual requests. ### Provider benchmarks and market data The service says it regularly tests providers with standard benchmarks, publishes results, and provides live and historical prices for models. ### Provider marketplace tools Model providers can register deployments, set and adjust prices, control capacity, and review jobs served, win rates, auction losses, earnings, and payouts. ## Use Cases ### Cost-aware application inference Route production or development requests through eligible model providers while using request requirements and caps to manage inference spend. ### Coding agents and terminal workflows Use Liquid CLI with supported terminal agents such as Claude Code, Codex, or OpenCode, selecting a ranking strategy and request cap. ### Existing API client migration Connect applications that already speak the OpenAI or Anthropic API by changing the client base URL and supplying a Liquid Inference API key. ### Inference capacity monetization Operators with model deployments can list them on the exchange, quote prices based on available capacity, and receive payment for tokens served on won requests. ## Frequently Asked Questions ### How does Liquid Inference choose a provider? You set routing terms such as models, regions, speed, and minimum quality. Providers that meet those terms compete, and the lowest-priced qualifying offer answers the request. ### What does a buyer pay for? The site states that buyers pay for the tokens they use. Its pricing display separates output-token pricing from input-token pricing, and itemized billing shows request costs. ### Can I limit the cost of a request? Yes. Liquid CLI supports a cap, and the service describes prices as capped before the model starts when a provider is selected. ### Can existing AI tools connect to Liquid Inference? The service documents a base-URL change for clients compatible with the OpenAI or Anthropic APIs and lists integrations including coding agents, SDKs, and workflow tools. ### How can a model provider join? A provider creates an account, registers a deployment, waits for an operator review and listing, then connects an agent to begin quoting. The site says providers pay no listing or transaction fees. ## Quick Facts - **Category**: LLM inference routing and marketplace - **Provider**: Architect - **Primary users**: AI application teams, developers, and model providers - **Buyer interface**: API, Liquid CLI, and Liquid MCP - **Routing model**: Per-request auction based on buyer-defined requirements - **Website**: liquidinference.ai