Kunlun Chip
Kunlun Chip is a series of AI chips independently developed by Baidu, designed for deep learning training and inference workloads. The chips target cloud and edge AI computing requirements, supporting tasks such as natural language processing and image recognition. Key characteristics include high computational throughput and low power consumption. The chips are compatible with multiple deep learning frameworks and form a core part of Baidu's strategy for self-controlled AI infrastructure.
Timeline (last 90 days)
For other domestic chip vendors (Cambricon, Hygon, Kunlunxin, Moore Threads, MetaX, Biren, Enflame, and Iluvatar CoreX) to fully adapt DeepInfer, they would roughly need to overturn their existing engines and fill in about 30 TileLang backend operators, FP8×FP4 GEMM fusion, Mega MoE, sparse attention, and TopK components.
A total of 10 chipmakers—including domestic chips such as Ascend and Kunlunxin, as well as NVIDIA and AMD—announced that they are integrating MiniMax's open-source model.
In "chip-cloud-model-agent," chip refers to self-developed AI chips such as Kunlun chips; cloud refers to Baidu AI Cloud; model refers to the ERNIE series of large models and their toolchain; and agent refers to the agent layer.
Baidu's "Chip-Cloud-Model-Agent" framework corresponds to four levels: Kunlun Chip AI chips, Baidu AI Cloud, the ERNIE large model series, and intelligent agents and end-user application products.
All Facts (5)
LoneForge支持同一套代码在GPU和昆仑芯等不同硬件平台上运行
60%UnverifiedFor other domestic chip vendors (Cambricon, Hygon, Kunlunxin, Moore Threads, MetaX, Biren, Enflame, and Iluvatar CoreX) to fully adapt DeepInfer, they would roughly need to overturn their existing engines and fill in about 30 TileLang backend operators, FP8×FP4 GEMM fusion, Mega MoE, sparse attention, and TopK components.
50%UnverifiedA total of 10 chipmakers—including domestic chips such as Ascend and Kunlunxin, as well as NVIDIA and AMD—announced that they are integrating MiniMax's open-source model.
50%UnverifiedIn "chip-cloud-model-agent," chip refers to self-developed AI chips such as Kunlun chips; cloud refers to Baidu AI Cloud; model refers to the ERNIE series of large models and their toolchain; and agent refers to the agent layer.
50%UnverifiedBaidu's "Chip-Cloud-Model-Agent" framework corresponds to four levels: Kunlun Chip AI chips, Baidu AI Cloud, the ERNIE large model series, and intelligent agents and end-user application products.
50%