Find lower-cost options for text, images, video, embeddings and APIs. Compare published rates, bring your own keys, or use your own GPU.
OpenRouter text, image and embedding requests now prioritize provider price. Your selected model stays the same. Availability, provider fees and latency still vary; direct-provider and video requests retain their native pricing.
Compare live token estimates before running a prompt.
Compare model pricing and use a smaller supported size for drafts.
Start with short clips and review supported resolution and duration.
Compare input-token prices; avoid embedding unchanged text again.
Choose the required capability and check provider billing units.
Use your provider account directly and inspect usage after each run.
Review each service's published payment terms before authorizing.
Run installed models on your own computer without a hosted token bill.
Token estimates only. Models differ in quality and capabilities; changing models is always your choice. Media and request-based fees are not included. Provider rates apply; INFERDEX does not currently offer a discounted hosted tariff.
Balanced weights cost rank and context capacity, not answer quality. Fast uses your successful session requests, which may have different prompt lengths. No latency is invented.
Enter two published quotes in the same currency and billing unit. Compare the same resolution, duration, quality and service scope. These are your estimates, not INFERDEX prices.
Enter both prices to calculate. No assumed discounts.
Connect your own funded provider account and pay the provider directly for cloud requests. Compare costs here before you run, set spending limits with your provider, and review recorded usage in INFERDEX. Local inference uses your own hardware and electricity.
INFERDEX hosted credits and GPU rentals are not available for purchase. A connected wallet does not fund cloud requests. Savings vary by model, provider and workload; direct API rates remain the provider's rates.