# Agave v4.5 sets a date for Solana technology in December

> Anza publishes the Agave v4.5 schedule: mainnet activation on December 14, lower rent, direct access to accounts and two-step adoption.
> By José Maldonado · Blockchain · 2026-11-12

December 14, 2026 is the date Anza has set to begin activating on mainnet the features of Agave v4.5, the validator client used by most of the Solana network. The document lives in [the Agave v4.5 page on the Anza wiki](https://github.com/anza-xyz/agave/wiki/v4.5-Release-Schedule), updated on November 10, and strings together eight milestones between November 6 and that second week of December. No date is guaranteed: Anza itself repeats that they are tentative and that operators should wait for the Discord notices before touching a node.

For network users, the summary fits in two sentences: opening an account will cost less, and programs will no longer copy account data in order to read it. For those running a node, the important change is in the process: mainnet adoption goes from three steps to two, which cuts a week off the deployment cycle.

The calendar pauses between November 11 and 18 for Breakpoint, the Solana Foundation's annual conference held in London from the 15th to the 17th, and resumes on the 19th with the reactivation of functions across the three clusters. The testing window is split into two halves and the margin until December is measured in weeks.

## A calendar that reduces the steps

The v4.4 cycle called for three gestures (10%, 20 or 25% of the stake, and general adoption) and the v4.5 one calls for two. On November 30 Anza will call on volunteers with 20% of the stake, and on December 7 it will recommend the version to all validators. The team summed it up on social media as one week less per cycle.

That first step is not cosmetic. The [network validator data](https://solanacompass.com/news/anzas-agave-v45-schedule-cuts-solana-mainnet-adoption-to-two-steps-and-pauses-feature-activations-for-breakpoint) put activated stake at 432.4 million SOL spread across 658 validators in epoch 1083, so 20% is around 86 million SOL. A mid-sized validator that thinks twice leaves the step in the hands of the large operators, who respond to these calls sooner.

The v4.5 branch was created on November 5, the testnet received the recommendation on the 6th and the devnet was updated on the 10th. The mainnet candidate is tagged on November 27 and that same December 7 the v5.0 branch opens.

The immediate precedent argues against taking anything for granted. On November 5 Anza called for an unscheduled patch, Agave 4.4.1, before epoch 1079 arrived, and published it that afternoon. A cycle that already needed an emergency stop may need another.

## Byte rent falls in three stages

Solana's rent is a refundable deposit that each account holds per byte of data, not a payment that is consumed. The proposal [SIMD-0437](https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0437-incremental-rent-reduction.md) sets out to lower it from 6,960 lamports per byte to 696 in five steps, and Anza includes the last three in this cycle. The current price, 5,080 lamports per byte, drops to 2,575, then to 1,322 and finally to 696.

A standard token account takes up 165 bytes and the system adds 128 bytes of overhead, so the deposit falls from about 0.0015 SOL to about 0.0002 SOL once all three phases are active. The first two steps have been tested on testnet since August and the total reduction reaches 90% of the original bill.

Accounts that already exist continue to operate with the old deposit and [their holder can withdraw the surplus after each phase](https://solanacompass.com/news/anzas-agave-v45-schedule-cuts-solana-mainnet-adoption-to-two-steps-and-pauses-feature-activations-for-breakpoint)It is the least flashy part of the announcement and the one that shows up most in a portfolio with hundreds of small accounts.

A deposit seven times smaller changes the bottom line of an app that creates one account per user, and also that of airdrops that distribute balances to tens of thousands of new addresses.

## Programs read the account where it is

The current runtime copies the account data into the program's memory and returns it when finished, and repeats that copy on every call between programs. [SIMD-0460](https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0460-virtual-address-space-adjustments.md) removes the intermediate step: the program reads and writes on the data where it is, including when it calls another program.

Jonas Hahn, from the Solana Foundation developer team, asked on November 10 to test programs with the virtual\_address\_space\_adjustments feature enabled, locally or against devnet, because some still show divergences with the new mapping.

That detail matters to anyone running code in production: a program that assumed it had its own copy of the data can behave differently when that copy disappears. Testing on devnet stops being a formality and becomes the only way to verify it.

## The upgrades adjust the program size

Solana's program loader allows data to be expanded with the ExtendProgram instruction, but not reduced, which is why a project that slims down its binary keeps paying rent for bytes it no longer uses. [SIMD-0433](https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0433-loader-v3-set-program-data-to-elf-length.md) addresses both halves of the problem.

With the change, the upgrade resizes the account to the size of the new binary: a larger program no longer needs the prior step, and a smaller one returns the surplus rent to the account the developer specifies. Frequent deployers save on the deployment script and on the rent they stop paying for empty bytes.

## The timeline, milestone by milestone

The dates summarizing the cycle have been on the wiki since November 10. The first two rows have already happened; the rest are targets that Anza can move with no notice beyond an edit to the page.

Milestone

Target date

Status

Branch v4.5 created

November 5

completed

Testnet recommends v4.5

November 6

completed

Mainnet candidate tagged

November 27

pending

Volunteers with 20% of stake

November 30

pending

General adoption recommended

December 7

pending

Feature activation on mainnet

December 14

pending

Two milestones worth keeping in mind fall outside the table: the reactivation of features on testnet, devnet and mainnet on November 19, and the devnet update on the 10th. Both serve as a rehearsal before the code touches real money.

## Breakpoint splits the cycle in two

The Breakpoint pause is not a scheduling detail. Between November 11 and 18 Anza freezes release activity, and when it returns on the 19th it has to reactivate features across all three clusters at once. Less than four weeks separate that resumption from the final activation on December 14.

The margin narrows because the client has just been through a consensus change. Alpenglow was activated on mainnet on October 21 with [Votor and a 150-millisecond finality](https://observatorioblockchain.com/en/blockchain/solana-alpenglow-consenso-150ms-finalidad/), compared with the previous 12.8 seconds, and required validators to update the client.

The three v4.5 fronts share a runtime, so a rent failure can show up as a direct mapping failure and vice versa. The Breakpoint pause, moreover, bundles the tests into the same November window instead of spreading them out.

The 20% of the stake does not move on its own: operators who answer to their users have to be convinced. When a staking provider fails, the effect reaches the depositor before it reaches the network, as happened with [Solflare's reward freeze](https://observatorioblockchain.com/en/defi/solflare-staking-proveedor-incidente-recompensas/) at the end of October.

## What Solana has at stake in five weeks

Anza has spent two cycles trying to make mainnet adoption less ceremonial. The 10% threshold required two rounds of coordination and two waits, and the result was the same: everyone upgraded. Removing it cuts the time and the chances of catching a problem before it reaches everyone.

Execution risk is concentrated on November 30. If volunteers do not reach 20% of stake, the schedule stretches and December 14 stops being a date. Large operators respond quickly because they have a team; the small validator who alternates between [Firedancer](https://observatorioblockchain.com/blockchain/solana-y-firedancer-buscan-eficiencia-web2-para-atraer-inversores/) and Agave decides with less information.

For the developer, the useful part of the announcement is not the dates but the three features: cheaper accounts, data without copying and updates that adjust themselves. For the validator, the useful part is the notice that there is a window on devnet and another on testnet before touching mainnet.

The December 14 date will move if something goes wrong, as others have moved this cycle. What does not move is the direction: less deposit per byte, fewer copies in memory and fewer manual steps in updates.
