Gonka Network H100/H200 and other high graphics memory devices have a computing power of over 9000 cards and support nearly 40 GPU hardware models Odaily Planet Daily News: According to the latest data from Gonkascan. com, Gonka's network computing power has exceeded 12000 H100 equivalent images, and the overall scale is close to the level of traditional large-scale AI computing centers. With over 9000 high memory GPUs such as H100/H200, the proportion continues to increase, leading the effective computing power weight of the entire network and enabling the network to stably carry high concurrency inference tasks with billions of parameter models. The widespread participation of mid to high end graphics cards has significantly reduced the threshold for node access, as long as the minimum video memory requirements are met, they can participate in MLNode operation. Data shows that Gonka Network supports nearly 40 GPU models, covering data center level computing power such as H200, H100, A100, as well as professional and consumer grade graphics cards such as RTX 6000 Ada, A6000, A40, RTX 4090, forming a multi-level, composable global computing power supply structure, reflecting Gonka's highly compatible and efficient decentralized computing power path, providing more open and flexible infrastructure support for AI inference and training. At the same time, the five AI inference models of Gonka Network have exceeded 3000 daily users, and the demand for inference continues to grow at a high level. The model represented by Qwen3-235B-A22B-Interstructure-FP8 has formed continuous and high-frequency stable calls in the Gonka network, and its usage form has clearly deviated from the testing stage, approaching the early actual production load of mainstream AI APIs. Gonka is committed to efficiently integrating global GPU computing power through a decentralized approach, providing permissionless infrastructure for AI inference and training. The project is incubated by American AI developer Product Science Inc., with the founding team consisting of Libermans and his siblings, former core product leaders of Snap Inc. The total financing amount exceeds $69 million, and investors include Coatue (OpenAI investor), Slow Ventures (Solana investor), Bitfury, K5, Insight, Benchmark partners, and others.