About Cheap Tokens

Open-model inference at a tenth of the list price, run on idle GPUs in your region.

Why it exists

Per-token prices for open-weight models are still set by datacenter costs, so a small app doing thousands of summaries a day pays a bill that scales linearly with its success. Cheaper relays exist, but they resell closed models through unclear arrangements, publish latency claims with no per-request proof, and give the buyer no way to see where a prompt ran or to get money back when a call is slow.

What it does

• One-key gateway
• Host console
• Job receipts
• Prepaid balance and plans
• Key spend caps and regions