The core model of these markets primarily revolves around discounted reselling of API call quotas and idle inference compute power, with some projects building networks that enable demand-side users to directly connect with independent compute service providers.

In this emerging sector, three representative markets stand out: Orbio deployed on Robinhood Chain, and Surplus Intelligence and Antseed deployed on Base. Although all three offer low-cost inference services, they exhibit clear differences in service positioning, target users, and token economics. Below, we conduct a horizontal comparison of these three projects.

Orbio

Orbio’s inference marketplace is primarily designed for AI agents. Originally just a token distributing OpenRouter quotas to holders, it has evolved into a full-fledged protocol supporting hundreds of AI models with significant price discounts. It also features a Launchpad where developers can deploy agents and issue tokens paired with ORBIO for trading.

To better serve agent-specific use cases, Orbio integrates privacy-preserving computation powered by Phala and NEAR AI. Recently, it partnered with Mosh, incorporating Mosh’s tooling ecosystem to help new agent tokens establish more sustainable token growth mechanisms.

These capabilities collectively empower agents on Orbio with greater autonomy—enabling them to invoke inference compute, manage funds, and even autonomously execute more on-chain actions.

The cornerstone of this model is the CREDIT voucher: one CREDIT represents $1 worth of inference capacity within the Orbio gateway. Agents can purchase and burn CREDIT using wallet funds to access compute power, ensuring sustained operation. Tokenizing inference capacity allows agents to autonomously procure compute for survival, often at discounted rates.

For example: CREDIT’s recent market price is approximately $0.47. An agent spends $47 to buy 100 CREDITs, and after fees, can redeem them for $95 worth of inference capacity.

The discount originates from ORBIO’s transaction fees, which are converted into CREDIT rewards distributed to users. Reward recipients can then sell these CREDITs on secondary markets. Thus, higher trading volume leads to more CREDIT rewards; additionally, smart contract tokens launched via the Launchpad and traded against ORBIO form their own closed-loop circulation.

The question remains: can this model sustain itself without speculative behavior? Although the platform has disclosed data since September, on-chain analytics show that the vast majority of issued CREDITs remain unused to date.

Surplus Intelligence

Among the three, Surplus Intelligence’s model is the most straightforward—an AI compute aggregation marketplace at its core. Built on the Base chain, service providers can list interfaces for models such as OpenAI, Anthropic, Venice, and OpenRouter, setting their own prices. Market competition continuously drives down pricing.

Upon receiving inference requests, the platform routes them according to predefined rules to corresponding service providers. Buyers can pay via credit card or USDC, and agents can also integrate and invoke services directly. This simple, direct model has made Surplus the most actively used among the three platforms. Surplus reports having processed a large volume of inference requests, with many calls costing less than $0.01.

Beyond low cost, Surplus has introduced new features to enhance competitiveness. Recently, it launched a key obfuscation feature: once enabled, the system automatically replaces sensitive information like API keys and private keys with placeholders before forwarding prompts to model providers. Providers only see placeholders, and the original data is restored post-response. This feature is highly beneficial for agents handling keys, although the platform itself still reads the original prompt.

Despite high usage, Surplus’s product and token linkage is the weakest. The SURPLUS token cannot be used to pay for inference, does not participate in compute routing, and lacks a public mechanism to distribute platform revenue to token holders. Future changes may occur—team members are advancing a token governance rights adjustment plan—but nothing has been implemented yet. In short, while the market is operational and activity continues rising, increased usage currently fails to directly translate into token value.

Antseed

Where Orbio focuses on agent-centric scenarios and Surplus emphasizes extreme pricing, Antseed takes a different path—it brands itself as “BitTorrent for AI inference.”

Rather than relying on a centralized gateway to forward requests, Antseed builds an open compute discovery network that integrates various AI models. Service providers can independently deploy their compute resources. Buyers autonomously search for available providers, connect through routing APIs, and freely select compute suppliers.

This model offers superior flexibility. The network is open to all providers, resulting in highly diverse supply: from affordable, popular models to specialized model validation services verifying whether providers genuinely deliver the claimed models. Buyers can choose based on their specific needs rather than being constrained by centralized gateways allocating resources.

The project’s native token, ANTS, is currently not publicly tradable. Its token economy is centered around a reputation system within this market. Both buyers and sellers can stake tokens—providers staking ANTS increase their reputation and qualify for rewards.

Differently from Orbio, reward funding does not come from transaction fees. Instead, Antseed funds rewards from treasury reserves. While future revenue streams may be opened, no such mechanisms have been implemented yet. Therefore, the positive correlation between growing inference demand and token value remains unverified at present.

Summary

Notably, the differentiation among these three projects goes beyond simply competing on price: Orbio targets the agent ecosystem, enabling agents to autonomously acquire—or even self-fund—their compute; Surplus focuses on achieving the lowest possible costs, enhanced by practical features like key obfuscation; Antseed establishes an open P2P compute market where users can select providers based on their exact needs.

As projects continue iterating, additional features will likely emerge to attract users. Will it be a novel compute revenue model? Or a comprehensive ecosystem tailored for agent development? We await further developments.

Written by: David Christopher, Translated by: Chopper, Foresight News