MiMo-V2.6-Pro ties Grok 4.7 (xHigh) and beats GLM-5.3 (max) on Artificial Analysis' Intelligence Index, making it the benchmark's top-scoring open-weight model

The top open-weight models now match proprietary performance benchmarks.

MiMo-V2.6-Pro, an open-weight model from Xiaomi, has achieved the top score among open-weight models on the Artificial Analysis Intelligence Index, scoring 46. This places it significantly above the median score of 18 for comparable models. The model excels in intelligence and speed, processing 130 tokens per second, which is notably fast. It supports multimodal inputs including text, image, speech, and video, and boasts a 1M token context window. While it demonstrates strong performance, it is described as somewhat verbose and its cost per input token is considered slightly expensive compared to its peers. MiMo-V2.6-Pro also performed better than Grok 4.7 (xHigh) and GLM-5.3 (max) on this specific intelligence benchmark. The model's weights are available under an MIT license.

MiMo-V2.6-Pro's performance on the Artificial Analysis Intelligence Index signals a narrowing gap between open-weight and proprietary large language models. Its ability to achieve a score of 46, outperforming models like Grok and GLM, suggests that the open-source community is rapidly closing the performance deficit. This advancement will likely accelerate innovation by making highly capable models more accessible for researchers and developers to build upon and fine-tune for specific applications.

The multimodal capabilities and large context window of MiMo-V2.6-Pro, coupled with its strong intelligence score, indicate a shift towards more versatile and powerful open-weight models. This increased accessibility of advanced AI technology could democratize AI development, potentially leading to a wider array of specialized AI applications and services. Developers and businesses can now leverage top-tier intelligence and multimodal processing without the costs or restrictions associated with proprietary systems, fostering greater competition and faster iteration in the AI landscape.