Chain Future

Solidity, smart-contract development, developer tools and explained protocol engineering

Pearl’s FP8 Testnet Update: What It Changes for AI-Inference Mining

For an AI-inference provider considering Pearl, the practical question is whether work it already performs for customers can also qualify for mining rewards. In a report published on November 19, 2026, Bankless analyst David Christopher described Pearl’s FP8 upgrade as a step toward that goal. The upgrade went live on testnet on November 18; the report did not establish that providers had adopted it.

Pearl is a fork of Bitcoin’s code. Its proposed “Duplex Work” mechanism would let providers submit proofs of AI-inference computations to earn block rewards. But Christopher noted a central limitation: miners were not required to perform real AI inference and could still earn tokens by computing meaningless blocks.

Pearl’s FP8 Testnet Update: What It Changes for AI-Inference Mining

What FP8 changes

Under the previous mechanism, qualifying work had to use the INT7 format. That meant a provider using another format would need to convert its model before participating. The testnet upgrade moves the mining standard to FP8, a format the report says is widely used by commercial inference platforms. It also claims that more than 80% of model-inference traffic on platforms such as OpenRouter runs in FP8. If providers can use their existing workloads, that could reduce the work needed to connect them to Pearl.

The format change has a hardware trade-off: according to the report, INT7 can run on a broader range of older graphics cards, while FP8 typically requires newer GPU architectures. Nor does FP8 alone prove that a submitted workload came from a real customer request. The report says more sophisticated on-chain verification would still be needed to address that problem.

Christopher identified three tests for the project’s premise: whether established AI providers publicly say they mine PEARL and pass token subsidies on through cheaper API access; whether mining power becomes less concentrated; and what share of blocks represents genuine user inference. He noted that the last measure is especially difficult to observe externally.

Crypto researcher hitesh.eth offered a more skeptical assessment. He argued that the case for Pearl depends on Bitcoin miners moving into AI computing for token rewards, and warned that high token inflation could make such incentives unsustainable. He also alleged coordinated promotion around the upgrade and doubted that a higher token price would translate into real product adoption. Those are his criticisms, not outcomes demonstrated by the testnet release.