Moore Threads has teased its upcoming Lushan graphics card, which the company claims will feature the largest video memory capacity in the consumer market. The official launch is scheduled for December 29, 2026, during the MUSA Developer Conference. The manufacturer intends the card for both high-end gaming and local artificial intelligence model execution.
While the exact memory capacity remains undisclosed, the company suggests it will surpass the 32 GB found in the GeForce RTX 5090. This hardware configuration is primarily aimed at running local AI models, such as Qwen3.8-27B, which requires 27–30 GB of VRAM for 8-bit operations. The card also relies on memory bandwidth to handle these intensive computational tasks.
The Lushan card introduces the new UNITE rendering architecture, a next-generation hardware ray tracing engine, and support for AI-driven generative rendering. Moore Threads claims the device will offer up to 64 times the AI performance of its previous models. The company has previously demonstrated significant performance gains through driver updates, including doubling the speed of the MTT S80.
The company's track record includes the MTT S90, which outperformed the GeForce RTX 4060 in earlier tests. Despite these improvements, the actual market competitiveness of the Lushan card will depend on independent performance benchmarks following its release. The product represents a strategic shift toward combining gaming capabilities with specialized local AI processing power.





