Surplus Intelligence

Surplus Intelligence is a decentralized exchange for LLM inference. Sellers list spare OpenAI-compatible capacity at their own price; buyers use one API key and base URL to get routed to the cheapest offer. Settlement is non-custodial USDC on Base, with three rails: traditional API keys, x402 micropayments, and MPP (Tempo). An Anthropic-compatible layer lets Claude Code route through the marketplace.

Its core value is price discovery for unused model capacity, plus agent-native payments: x402 enables per-request USDC settlement without an account. The platform runs at zero take rate (fee multiplier 1.0x) to build liquidity. The pain point it addresses: unused API credits or committed capacity expire worthless, and coding agents face high per-call costs. It solves this by turning surplus capacity into a sellable asset and routing buyers to the lowest listed price, claiming a 38% cost cut (self-reported).

From Mar 13 to Sep 13, 2026: the marketplace grew to 145 catalog models and 76 active market listings (snapshot). 28-day usage reached 31.17M requests, 1.84T input tokens, and 28.98B output tokens. 7-day realized pricing showed an 82% mean discount vs direct provider prices. Seller breadth remains thin: 11 allowlisted domains, with active offers concentrated on Venice and Bankr LLM Gateway.

Negative aspects within two years: no cryptographic proof that a seller served the exact claimed model, and no automatic detection of proxying to cheaper hidden models. The settlement contract has no timelock, so an admin can change fees in one transaction. A permanently failed settlement carries no automatic seller reimbursement. The SURPLUS token has no documented role in fees, staking, or routing.