The upgrade proposal for the next on-chain software version v0.2.12 has now been published on-chain and is open for voting.
Key changes
— Multi-model PoC (the largest change) (#1039). Transition Proof of Compute from a single fixed model to per-model PoC groups. Each governance-approved model generates its own local PoC weight, which is then aggregated into a total consensus weight via model-specific coefficients. Each host must participate in each model group (either directly or by delegating PoC voting weight).
— moonshotai/Kimi-K2.6 is introduced as the second model: The model group will be activated two epochs after the upgrade. The coefficient for this model is 3.51x the coefficient of Qwen235B, based on compute complexity of models on the same hardware (8xH200, 8xB200).
— Devshard standalone runtime (#1045). Decouples devshard releases from the DAPI / mainnet release cycle.
— Certik audit fixes. Audit findings have been addressed.
— Protocol hardening. Preserved nodes (POC_SLOT=true are randomly sampled for single PoC / CPoC time. Other updates include propagating the mlnode version to the on-chain HardwareNode, fixing DKG dealer consensus, aligning legacy validator slashing with required-collateral semantics, ensuring atomicity of the devshard escrow fund, and adding zero-timestamp tolerance to inference_finished event parsing.
Upgrade plan
The binary versions will be updated via an on-chain upgrade proposal. For more information on the upgrade process, refer to /docs/upgrades.md.
Required actions
1. Before upgrade
Deploy latest versions of versiond and proxy services from docker-compose.yml (using the repo at tag release/v0.2.12):
shell
git checkout release/v0.2.12
Deploy (important to use --no-deps):
shell
source config.env && \
docker compose -f docker-compose.yml up versiond proxy -d --no-deps
That will activate devshard working independently from api service.
2. Post upgrade
Deploy, delegate, or explicitly refuse new governance-approved model(the included model will be activated 2 epochs after the upgrade). Refer to the guide.
3. Before or after upgrade
Hosts are asked to update the dashboard/explorer. Please run the following commands from the gonka/deploy/join directory:
shell
docker compose -f docker-compose.mlnode.yml -f docker-compose.yml pull explorer
docker compose -f docker-compose.mlnode.yml -f docker-compose.yml up -d explorer
How to vote
If you do not have direct access to the key that holds voting power, or want another key to vote on your behalf, please refer to the guide on granting governance voting permission from a cold key to a warm key.
Proposal details and voting are available via inferenced. Any active node can be used. Available nodes include:
— http://node1.gonka.ai:8000
— http://node2.gonka.ai:8000
— https://node3.gonka.ai
Cast your vote ( yes, no , abstain , no_with_veto ):
shell
export NODE_URL=https://node3.gonka.ai/
./inferenced tx gov vote 44 yes \
--from <cold_key_name> \
--keyring-backend file \
--unordered \
--timeout-duration=60s --gas=2000000 --gas-adjustment=5.0 \
--node $NODE_URL/chain-rpc/ \
--chain-id gonka-mainnet \
--yes
To check the voting status:
shell
export NODE_URL=https://node3.gonka.ai/
./inferenced query gov votes 44 -o json --node $NODE_URL/chain-rpc/
Deadlines
Voting ends: April 30th, 2026, at 00:12 UTC
Upgrade height: 3834200
Estimated upgrade time: April 30th, 2026, at 6:00 UTC
Attention
— Please plan to be online during the upgrade window so that any follow-up steps or mitigation instructions can be applied promptly.
— During upgrades, Cosmovisor creates a full state backup in the .inference/data directory; ensure sufficient disk space is available. Guidance on safely removing old backups from the .inference directory is available in the documentation.
— If application.db occupies a significant amount of disk space, the cleanup techniques described in the cosmovisor backup guide may be applied.
— After the upgrade, Postgres is available as an option for local payload storage.