Introducing Liquid Inference by Architect
We're introducing Liquid Inference, the world's first dynamically-priced inference exchange.
Liquid Inference comes from the team's experience building exchanges and trading algorithms. We believe that the inference industry, similar to various financial markets, trades and prices token outputs dynamically. At Architect, we're determined to build the Exchange for AI industry and with that in mind, decided to build Liquid Inference as a platform for our users.
We announced the launch of Compute Futures on our US-exchange in May 2026 and since then have seen the AI industry and its market participants evolve dramatically. Whilst working with them, it became clear that token outputs were a measure of intelligence and productivity that enterprises were benchmarking against when making internal AI decisions. We wanted to meet our customers where they are today.
Why do you need Liquid Inference?
The inference market today is made up of large inference providers who focus on bilateral enterprise-grade contracts. Buying additional credits for volume AI-work requires either purchasing credits directly from the labs or using markets like OpenRouter. However, suppliers aren't always incentivized to offer their "best-bid" on these platforms and especially not real-time.
We know that token usage has peak and off-peak hours and so does the underlying operating expense, power. Ollama, an open-source platform to run LLMs locally on your computer already offers cheaper off-peak pricing on token spent before 12pm and after 6pm.
Liquid Inference's dynamic pricing will enable fair competition for inference which may differ for standard daily tasks, low latency tasks and high throughput workloads. We aim to be the price discovery platform for all of the above.
Liquid Inference is fully compatible with agentic coding tools like Claude Code, Codex, OpenCode, Cursor, Pi, Cline and many more. The exchange will offer hundreds of open- and closed-weight models and users can create custom routing rule presets or use Architect's Auto routing algorithms.
Partner Network
We're excited to be working with partner inference providers who are bootstrapping the network in the platform's early days with us. They bring a vast expertise in managing compute and inference workloads and are in the first cohort of inference providers on the platform. We're seeing hundreds of tasks completed successfully in the test phase of the product across 700+ models.
New providers can sign-up and start quoting in minutes, not weeks. All prompts are compatible OpenAI API standards and only need a REST/Websocket API to register models and quotes. Payouts are processed via Stripe.
Join Now!
- To Users: We're offering $20 in credits to the first 500 users. Early users will receive 20% of Architect fees in credits for referred users and 10% for referrals of referrals.
- To Providers: No fees charged to the providers, free distribution and simple payouts via Stripe.
- URL: liquidinference.ai
- Contact: sales@architect.co
- Join our community: Discord
Liquid Inference is a service provided by Architect Financial Technologies Inc., a Delaware corporation ("Architect"). Liquid Inference is an AI inference marketplace. Users purchase inference services from Architect. Architect sources inference from independent providers, which act as its subcontractors and have no contract with buyers. Liquid Inference is not a financial, investment, or digital asset product. Use of the Liquid Inference service is subject to the Terms of Service and Privacy Policy available at inference.ai.exchange. It is not available in all jurisdictions. Credits purchased through Liquid Inference are prepayments for services provided by Architect through its subcontractors and may be used only for those services. Credits are not transferable, have no cash value, and are not currency, a bank deposit, or a digital asset except as otherwise required by law. Model names are trademarks of their respective owners and are used only to identify models available through Liquid Inference. Their use does not imply endorsement by or affiliation with those owners. AI-generated outputs may be inaccurate, incomplete, or inappropriate. Review outputs before relying on them. Architect does not provide legal, financial, medical, or other professional advice.