The Solana mainnet has enabled the p-token upgrade. This update, based on the Pinocchio Rust library, replaces the existing implementation of the SPL Token program while maintaining compatibility with current wallets, applications, and tokens. According to data disclosed by the Solana Foundation, computation costs for token-related instructions can be reduced by up to 98%.
Transfer instruction volume has significantly decreased
The most notable aspect of this upgrade is the significant reduction in token transfer costs. The computational units required for a standard token transfer have dropped from 4,645 to 76, and the transfer_checked instruction has decreased from 6,200 to 105.
On the Solana network, compute units measure the resources consumed by transactions and program executions. The token program is one of the most frequently used on-chain components, with DeFi swaps, NFT trades, payments, staking, and stablecoin transfers all invoking its instructions. Therefore, optimizing the token layer does not just affect a single function—it impacts a broad range of on-chain activities.
The article mentions that switching to p-tokens alone could free up approximately 12% to 13% of block space. This estimate does not include any additional improvements that may result from future batch processing features.
Keep the standard unchanged; directly replace the old implementation.
This upgrade did not introduce a new token standard but was launched as a direct replacement for the existing SPL Token program. Developers do not need to migrate assets, and users do not need to take any additional actions—existing tokens will continue to function as usual.
The focus of this approach is to reduce upgrade friction. If compatibility remains stable, validators can fit more transactions into a single block without increasing the current block limit, which also helps improve transaction ordering efficiency.
Main optimizations come from entry points and reading methods.
According to Anza engineer Fernando Otero, approximately 70% of the computational savings come from two adjustments: replacing the traditional solana-program entry point and adopting a zero-copy reading approach.
The remaining optimizations stem from granular code-level improvements, including the removal of redundant checks, reduction of unnecessary borrow tracking, optimization of conditional logic, and prioritization of the most frequently used instructions. Since transfer instructions account for approximately half of the mainnet token program traffic, the engineering team specifically implemented early identification and streamlined parsing for these instructions.
In addition to improved efficiency, p-token now supports three new instructions: Batch, WithdrawExcessLamports, and UnwrapLamports.
Testing and fixes have been completed prior to launch.
Because this update rewrote one of Solana’s most critical programs, the team conducted multiple rounds of testing before launching on mainnet. Anza stated that the testing methods included unit tests, fuzz testing, historical transaction replay analysis, audits, and ongoing formal verification.
Asymmetric Research also discovered a vulnerability related to batch instruction account ownership checks during development. The Anza team stated that this issue was patched before mainnet launch.
Over the past year, external attention has focused heavily on Firedancer, Alpenglow, and transaction format upgrades; however, from a practical usage perspective, many developers view the p-token upgrade as a more direct performance improvement. In 2025, developers anticipated that this upgrade would increase the efficiency of common token operations by approximately 19 times, and mainnet benchmark results have since shown that some instructions have improved even beyond those early expectations.