Odaily Planet News: The decentralized AI computing power network Gonka has completed its mainnet upgrade to version v0.2.9. This upgrade was passed through on-chain governance voting and officially executed at block height 2451000. The network has now fully transitioned to PoC v2 as the weight allocation mechanism, while the original PoC logic has been simultaneously phased out. This upgrade marks a significant step forward in the maturity of Gonka's computing power verification mechanism and network governance.
After the upgrade takes effect, the Confirmation PoC becomes the authoritative source for network results, further enhancing the verifiability and determinism of computational contributions. At the same time, the network enters a single-model operation phase, reducing heterogeneous computational noise by unifying the model and validation standards, and providing a more stable infrastructure environment for decentralized AI inference and training. Currently, only ML Nodes running Qwen/Qwen3-235B-A22B-Instruct-2507-FP8 and using PoC v2 compatible images can participate in weight calculations. The transition period from Epoch 158 to 159 will be the first complete operational phase after PoC v2 is enabled.
According to real-time data from GonkaScan as of February 2, 2026, the total network computing power of Gonka is approaching 14,000 H100-equivalent GPUs, demonstrating the scale of a national-level AI computing cluster. Compared to approximately 6,000 H100-equivalent GPUs at the time Bitfury announced a $50 million investment in early December 2025, the network's computing power has shown a monthly growth rate of about 52%, placing it at a leading level among similar decentralized computing networks in terms of growth pace.
From the perspective of computing power structure, high-end GPUs such as NVIDIA H100, H200, and A100 account for more than 80% of the total network computing power, demonstrating Gonka's significant advantages in aggregating and scheduling high-performance computing resources. Currently, the network nodes have covered approximately 20 countries and regions across Europe, Asia, the Middle East, and North America, laying a solid foundation for building a global, single-point-risk-resistant AI computing power infrastructure.