BAM Changelog — Sept 23-Oct 9

Action Required: Alpenglow Is Coming — Run Jito-Solana v4.3.0-jito.0 during transition period
Alpenglow mainnet activation is approaching in mid-October (per latest Solana Foundation / Anza information). BAM has been tested on Testnet and on the dedicated Alpenglow test cluster, including transaction and bundle traffic across Jito-Solana clients during activation, and Jito will keep operating a BAM node on the Alpenglow test cluster. The confirmed activation epoch, maintenance timing, minimum supported version, and restoration updates will be published through Discord announcements. Read the blog post.
What to expect during activation:
All mainnet BAM Nodes will be placed in Maintenance Mode during the days leading up to activation. Validators connected to BAM revert to ordinary Jito-Agave Block Engine mode. BAM-native features, MPP (Maker Priority Plugin) and BAM DBI (Direct Bundle Ingest), are unavailable during this period; Block Engine bundles continue to work as usual. After activation, BAM nodes are brought back online gradually while performance is monitored.
Validator actions before activation:
Upgrade to a supported Jito-Solana release: We currently recommend v4.3.0-jito.01 and will confirm the minimum supported version before activation.
Do not run a production validator built with dev-context-only-utils (DCOU): It will be forked off during activation.
Do not run FireBAM or FrankenBAM during the Alpenglow transition: After activation, use version 26.10 or later; 26.09 is not Alpenglow-ready. During the transition, we recommend switching over to v4.3.0-jito.1 of the Agave-based client.
During the transition window:
Validators, FullSend users, and standard Block Engine bundle users: No action required.
MPP users: Submit oracle updates via TPU or Block Engine.
DBI users: Use the legacy Block Engine bundle-submission path.
Current mainnet-beta minimum versions are unchanged:

Operator FAQ from the community:
Can I run Jito during the Alpenglow transition?
Yes. Run Jito Agave (jito-solana) as usual. Only Block Engine mode is enabled during the transition, not BAM. If your validator is normally connected to BAM it will gracefully fall back to Block Engine mode when BAM nodes enter Maintenance Mode, and BAM nodes are re-enabled once they are healthy after activation.
Which version will be required for the Alpenglow transition?
Anza and the Solana Foundation have not yet specified the mainnet transition version. Watch the validator-announcements channel for updates.
200ms Slots
The slot time reduction to 200ms will be active at Epoch 1053, approximately 15:00 UTC on Friday, October 9. Anyone running Jito Agave on mainnet is strongly encouraged to be on v4.3.0-jito.0, which has the best performance optimizations. To verify, run solana --version and confirm:
solana-cli 4.3.0 (src:c21f9074; feat:c9ad34d2, client:JitoLabs)
BAM's auction cadence is derived from slot length and will tighten automatically from 25ms to 20ms rounds. No BAM configuration change is required.
Releases
Jito-Solana v4.3.0-jito.0 (October 2) is the recommended mainnet release.
Jito-Solana v4.4.0-beta.0-jito (September 29) is the recommended testnet release.
FireBAM v26.09.5 (September 30) is the recommended mainnet and testnet release.
Network Adoption
BAM now secures 164M SOL across 385 connected validators, 37% of total network stake as of Epoch 1050.
FireBAM reached a new all-time high of 7.3M SOL across 12 connected validators, 4.6% of total BAM stake.
BAM celebrated its one-year mainnet anniversary on September 25.
Performance
Continued performance upgrades and optimizations to BAM's bundle ingestion pipeline have kept BAM's block packing and rewards competitive, with average rewards and compute units per block ranking in the top three across clients on daily, 5-day, and longer timeframes.
Measured across Epochs 1041 through 1050, FireBAM validators averaged 101.6 percent of the AgaveBAM cohort for compute units per block and 93.2 percent for rewards per block. FireBAM's skip rate was 0.036 percent, compared with 0.040 percent for AgaveBAM, with seven of ten epochs recording zero skipped FireBAM slots.
BAM blocks continue to be fuller than the network: 10.5 percent of AgaveBAM blocks were classified as small, compared with 13.4 percent network-wide and 17.2 percent for non-BAM jito-agave.
Performance work shipped or root-caused in this window:
A Linux kernel networking (BBR) behavior that delayed the first slot of some leader windows by 90–330ms was root-caused. A first-batch padding mitigation is live, and a kernel fix is in the TEE pipeline.
A periodic 7–17 percent slowdown of BAM TEE hosts caused by host memory compaction was root-caused and is being fixed with a host setting change.
A false-sharing fix delivered a roughly 20 percent latency cut in replay, alongside memory allocator improvements.
Block Engine bundles are now forwarded directly to nearby BAM nodes in every region.
Pruning of low-quality Block Engine search keys raised BAM tips per block by 15.5 percent.
BAM Registry
The BAM Registry has run on testnet for over two weeks without anomalous behavior. A jito-solana validator has connected end-to-end through registry discovery, with cached-list fallback and failover validated. Validator-side discovery is merged on jito-solana master. Mainnet enrollment is targeted for late October, and FireBAM integration follows TLS work currently in progress.
Attestations
BAM Attestations: Road to Transparency is now published on bam.dev (September 28), covering the attestation update our Engineering Lead gave at the Solana Validator Community Call. BAM nodes that fail attestation are excluded from the fleet, and BAM TEE guests provide no shell access by design.
Third-Party BAM Operators
Early cohort of Third Party Operators are currently being onboarded to BAM testnet, with the goal of coming online in the coming weeks.
Ecosystem
Preconfirmation revenue distribution to validators is live: validators are opted in automatically, and the first payouts arrive this month. BAM validators continue to earn 35% of preconfirmation revenue.
MPP now supports Transaction V1 and more than 64 markets per program, up to 256. PropAMMs route roughly 80 percent of their updates through MPP.
The Jito Bundle Explorer now shows bundles sent during BAM leader rotations.
JIP-39
JIP-39 changes are live. The BAM Boost subsidy formula has been updated with new effective-stake bands and a 2% pool-share ceiling, and subsidy claim windows were extended from 10 to 20 epochs. The Jito-operated, JitoSOL Prime validator was onboarded to the stake pool on October 5. Over Epochs 1041 through 1050, ~1,750 SOL in BAM Boost rewards was distributed to roughly 360 validators.(Historical earnings: For fiscal 2026 Q3 (period ended 2026-05-28), Micron Technology reported basic EPS of 41.97, diluted EPS of 41.4, and net income of USD 47.268 billion. In the prior-year period ended 2025-05-29, basic EPS was 4.79, diluted EPS was 4.75, and net income was USD 5.338 billion. For fiscal 2026 Q2 (period ended 2026-02-26), basic EPS was 16.91, diluted EPS was 16.68, and net income was USD 19.025 billion; in the prior-year period ended 2025-02-27, basic EPS was 3.1, diluted EPS was 3.08, and net income was USD 3.453 billion. Consensus expectations: The next upcoming consensus is for fiscal 2026 Q4, with an EPS estimate of 32.5625 and no revenue estimate available.)