> ## Content Index
> Fetch the complete content index at: https://nordisk-ai.ghost.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# Alibaba Launches Marco-o1 to Challenge OpenAI's Advanced AI Models
- URL: https://nordisk-ai.ghost.io/alibaba-launches-marco-o1-to-challenge-openai-s-advanced-ai-models/
- Published: 2024-12-20T19:09:15.000Z
- Updated: 2026-08-27T16:54:32.000Z
- Description: Alibaba takes the lead with Marco-o1 - an advanced AI model that challenges OpenAI's technology. With improved problem-solving ability and powerful translation technology, Marco-o1 is changing the game. Is this the beginning of a new era for AI development?
- Author: Anders Hagen
- Tags: AI Models, AI, aisweden, Alibaba, Marco-o1, machine learning, nordiskai, OpenAI, reasoning models, language models, #Migrated-1787841634512, #wp, #wp-post, #Import 2026-08-27 14:41, #Import 2026-09-07 13:41

The MarcoPolo Team, part of Alibaba International Digital Commerce, has introduced Marco-o1, an advanced large language model (LLM) designed to handle complex reasoning tasks and open-ended problem-solving.

This model uses techniques like Chain-of-Thought (CoT) fine-tuning, Monte Carlo Tree Search (MCTS), and novel reasoning strategies. Marco-o1 is specifically designed to tackle challenging real-world problems where traditional AI models often fall short.

[See It on arXiv](http://chrome-extension//efaidnbmnnnibpcajpcglclefindmkaj/https://arxiv.org/pdf/2411.14405?ref=nordisk-ai.ghost.io)

### Open Access to Marco-o1

Marco-o1 is now available to researchers and developers on GitHub and Hugging Face. The model is built on the Qwen2-7B-Instruct architecture and has been fine-tuned using datasets that include both open CoT data and proprietary synthetic data.

Key features include:

**Monte Carlo Tree Search (MCTS):** Lets the model explore multiple reasoning paths based on confidence scores.

**Reasoning action strategies:** Improves problem-solving ability by using different levels of granularity in actions.

### Improved Results and Language Translation

Marco-o1 has already demonstrated its strengths in testing. The model shows a 6.17% improvement in accuracy on the English MGSM dataset and 5.60% on the Chinese version. Marco-o1 also stands out in machine translation, where it manages to correctly interpret complex phrases and slang expressions. One example is the translation of "这个鞋拥有踩屎感" to "This shoe has a comfortable sole".

#### Strong Competition in the AI Market

The launch of Marco-o1 comes shortly after the Chinese AI research lab DeepSeek, backed by High-Flyer Capital Management, introduced its own reasoning model, DeepSeek-R1-Lite-Preview. Both Marco-o1 and DeepSeek-R1 are challengers to OpenAI's latest models, and performance tests show they're on par with OpenAI's o1-preview when it comes to logical and mathematical reasoning.

[See It on Hugging Face](https://huggingface.co/AIDC-AI/Marco-o1?ref=nordisk-ai.ghost.io)