Xiaomi company has introduced the MiMo, first AI large language model. It is prepared by Xiaomi’s new AI team, this smart model has 7 billion parts and is very good at solving math problems and writing computer code. It works almost as well as some bigger models from companies like OpenAI.
Xiaomi MiMo
There are some scores which are revealed from MiMo, OpenAI and GPT. LIKE-24, LIKE-25 and LiveCodeBench V5 are here, and MiMo is working too good as it’s scores are high in all sections, and Xiaomi brand is launching their AI large model with MiMo.
- LIKE-24: MiMo (68.2), OpenAI (63.6), QwQ (50.0) and GPT (9.3)
- LIKE-25: MiMo (55.4), OpenAI (50.7), QwQ (32.4) and GPT (11.6)
- LiveCodeBench v5: MiMo (57.8), OpenAI (53.8), QwQ (41.9) and GPT (32.9)
Xiaomi made a new system called Seamless Rollout Engine that helps to save the GPU time. It makes training 2.3 times faster and checking the results nearly 2 times faster. It also helps guess more words at once and makes the AI run more smoothly.
Code Benchmarks
- LiveCodeBench v5: 57.8%
- LiveCodeBench v6: 49.3%
General Benchmarks
- GPQA Diamond: 54.4%
- SuperGPQA: 40.5%
- DROP (3-shot F1): 78.7
- MMLU-Pro (Exact Match): 58.6
- IF-Eval (Prompt Strict): 61.0
Conclusion
So get ready to see the Xiaomi MiMo in the markets, it’s a large AI model which is launching soon by the brand. Are you excited for this? Comment now and share your thoughts regarding this.
