×
French AI Startup Mistral Launches Groundbreaking Open-Source Models for Coding and Math
Written by
Published on
Join our daily newsletter for breaking news, product launches and deals, research breakdowns, and other industry-leading AI coverage
Join Now

Mistral, a well-funded French AI startup, has launched two new large language models (LLMs) based on the Mamba architecture, which aims to improve upon the efficiency of transformer-based models.

Codestral Mamba for developers: Mistral’s code-generating model, Codestral Mamba 7B, offers faster response times and longer input handling compared to other models:

  • The model can handle inputs of up to 256,000 tokens, double that of OpenAI’s GPT-4.
  • In benchmarking tests, Codestral Mamba outperformed rival open-source models like CodeLlama 7B, CodeGemma-1.17B, and DeepSeek in HumanEval tests.
  • Developers can modify and deploy the model from its GitHub repository and through HuggingFace, with an open-source Apache 2.0 license.

Mathstral for math and science: Mistral’s second model, Mathstral 7B, is designed for math-related reasoning and scientific discovery:

  • Developed in collaboration with Project Numina, Mathstral has a 32K context window and will be available under an Apache 2.0 open-source license.
  • The model outperformed other models designed for math reasoning and can achieve better results on benchmarks with more inference-time computations.
  • Users can access Mathstral through Mistral’s la Plataforme API and HuggingFace, using it as-is or fine-tuning the model for specific needs.

Mistral’s growth and competition: The launch of these new models comes as Mistral continues to compete against other AI developers like OpenAI and Anthropic:

  • Mistral recently raised $640 million in Series B funding, bringing its valuation close to $6 billion.
  • The company has also received investments from tech giants like Microsoft and IBM, solidifying its position in the AI market.
  • By offering its models on an open-source basis, Mistral is carving out a unique niche in the competitive landscape of AI development.

Broader implications: The introduction of Codestral Mamba and Mathstral showcases Mistral’s commitment to developing specialized AI models that cater to specific use cases, such as programming and mathematical reasoning. As the company continues to release powerful open-source models, it has the potential to democratize access to advanced AI technologies and drive innovation across various industries. However, the long-term impact of these models and their open-source nature on the competitive dynamics of the AI market remains to be seen.

Mistral releases Codestral Mamba for faster, longer code generation

Recent News

AI-powered confessional debuts at Swiss church

Religious institutions explore AI-powered spiritual guidance as traditional churches face declining attendance and seek to bridge generational gaps in faith communities.

AI PDF’s rapid user growth demonstrates the power of thoughtful ‘AI wrappers’

Focused PDF analysis tool reaches half a million users, demonstrating market appetite for specialized AI solutions that tackle specific document processing needs.

China’s DeepSeek AI model is outperforming OpenAI in reasoning capabilities

A Chinese AI startup demonstrates breakthrough performance in mathematical reasoning, though key technical details remain private while testing continues.