Xiaomi MiMo

106 posts

Xiaomi MiMo

@XiaomiMiMo

Joined April 2025

17.2K Followers

Xiaomi MiMo’s posts

Pinned

Xiaomi MiMo-V2.5 is now officially open-sourced！ MIT License, supporting commercial deployment, continued training, and fine-tuning - no additional authorization required. Two models, both supporting a 1M-token context window : • MiMo-V2.5-Pro: built for complex agent and

Say Hi to MiMo-Audio! Our BREAKTHROUGH in general-purpose audio intelligence.

Scaling pretraining to 100M+ hours leads to EMERGENCE of few-shot generalization across diverse audio tasks!

Post-trained MiMo-Audio-7B-Instruct: • crushes benchmarks: SOTA on MMSU, MMAU,

MiMo‑VL 2508 is live! Same size, much smarter

We’ve upgraded performance, thinking control, and overall user experience.

Benchmark gains across image + video: MMMU 70.6, VideoMME 70.8. Consistent improvements across the board.

Thinking Control: toggle reasoning

Xiaomi MiMo-VL-7B 成功挑战2025年全国高考数学新课标Ⅰ卷，取得139/150的优异成绩。

MiMo-VL-7B 测试模型：huggingface.co/XiaomiMiMo/MiM，测试超参: temperature=0.3，top_p=0.95

多模态评测：MiMo-VL-7B & Qwen2.5-VL-7B

Today, MiMo can see We release MiMo-VL-7B-SFT and MiMo-VL-7B-RL, two powerful vision-language models delivering state-of-the-art performance in both general visual understanding and multimodal reasoning. MiMo-VL-7B-RL outperforms Qwen2.5-VL-7B on 35 out of 40 evaluated tasks,

Is it real or is it MiMo? Hear MiMo-Audio-7B-Base generate incredibly realistic and coherent dialogues for debates, podcasts, and more.

4:47

MiMo-Audio-7B-Base exhibits a very high "speech IQ" and strong modality alignment via few-shot examples.

See MiMo instantly detect and correct grammatical.

0:01 / 1:37

Replying to

Along side the remark MiMo-VL series, we also present MiMo-7B-RL-0530, which has seen significant improvements in reasoning and general capabilities through continuous reinforcement learning (RL) after the initial open-source release of MiMo-7B. In multiple mathematical coding

Today, Xiaomi releases MiMo, our first open-source reasoning model. At 7B parameters, it’s optimized for reasoning through pre-training and post-training, surpassing OpenAI’s o1-mini and QwQ-32B-Preview on AIME 2024-2025 and LiveCodeBench v5 benchmarks. #XiaomiMiMo #MiMo7B

Multilingual Mastery: Watch MiMo effortlessly switch between THREE languages for greetings & farewells.

0:01 / 0:42

Replying to

Thanks for sharing! We have a detailed introductory post here:

Quote

Xiaomi MiMo

@XiaomiMiMo

May 30, 2025

Today, MiMo can see We release MiMo-VL-7B-SFT and MiMo-VL-7B-RL, two powerful vision-language models delivering state-of-the-art performance in both general visual understanding and multimodal reasoning. MiMo-VL-7B-RL outperforms Qwen2.5-VL-7B on 35 out of 40 evaluated tasks,

Replying to

During the journey, we encountered two important observations that are worth sharing: (1) Incorporating high-quality, broad-coverage reasoning data from the pre-training stage is crucial for enhancing model performance. (2) Mixed On-policy Reinforcement Learning further

Replying to

Both the MiMo-7B-Base, SFT, RL-Zero and RL model checkpoints are now open-sourced and available at: huggingface.co/xiaomimimo For further details, please refer to the technical report:

MiMo/MiMo-7B-Technical-Report.pdf at main · XiaomiMiMo/MiMo

From github.com

Replying to

MiMo-7B-RL not only excels in code and algorithmic tasks but also outperforms both QwQ-32B-Preview and DeepSeek-R1-Distill-Qwen-7B across general tasks, even when the reinforcement learning evaluation is limited to mathematics and code problems.

Replying to

What can MiMO-VL-RL do?

In Case #1, our model showcases strong plot understanding capabilities, successfully converting an intricate plot into a well-structured markdown table. It also demonstrates superior reasoning capabilities in STEM tasks. In Case #2, MiMo-VL-7B

Replying to

On general benchmarks, the MiMo-VL-7B models, particularly MiMo-VL-7B-SFT and MiMo-VL-7B-RL, demonstrate consistently leading performance across a diverse range of vision-language benchmarks, surpassing other open-source models of comparable or larger scale.

Replying to

In the internal large model arena evaluating real user experiences, MiMo-VL-7B surpasses GPT-4o and Claude 3.7 Sonnet, becoming the top open-source VL model!

Replying to

On GUI understanding and grounding tasks, as a general-purpose VLM, MiMo-VL achieves comparable or even superior performance to GUI-specialized models, particularly on the more challenging Screenspot-Pro and OSWorld-G benchmarks.

Replying to

The comprehensive visual perception capabilities of MiMo-VL-7B are attributed to high-quality pre-training data and an innovative Mixed On-policy Reinforcement Learning (MORL) algorithm: Multi-Stage Pre-training: High-quality pre-training multimodal data has been collected,

Replying to

Thanks for sharing! We have a detailed introductory post here:

Quote

Xiaomi MiMo

@XiaomiMiMo

Apr 30, 2025

Today, Xiaomi releases MiMo, our first open-source reasoning model. At 7B parameters, it’s optimized for reasoning through pre-training and post-training, surpassing OpenAI’s o1-mini and QwQ-32B-Preview on AIME 2024-2025 and LiveCodeBench v5 benchmarks. #XiaomiMiMo #MiMo7B

Replying to

MiMo-7B-Base achieves significantly higher pass@k scores in reasoning benchmarks than all compared models, including the 32B baseline, across all benchmarks and evaluated k values. These results highlight the exceptional reasoning capabilities of MiMo-7B-Base.