During the Snapdragon Summit conference taking place this week in Hawaii, Qualcomm lifted the veil on a promising partnership with Mistral AI. This agreement aims to integrate the models of Large Language Models (LLM) recently published by the start-up, namely Ministerial 3B And Ministerial 8Bin devices equipped with the latest Snapdragon 8 Elite SoCs, as well as other advanced platforms aimed at the automotive sector.

Compact and powerful Mistral AI models

The models Ministerial 3B And Ministerial 8B stand out for their compactness while offering remarkable power. This architecture makes them perfectly suited for various devices such as:

• Smartphones
• Autonomous and connected vehicles
• PCs dedicated to artificial intelligence

Performance tests carried out by Mistral show that the model has 3 billion parameters outperforms its competitors, especially models Gemma 2 2B from Google and Llama 3.2 3B by Meta. As for the model equipped with 8 billion parametersit also outperforms Mistral 7B, Llama 3.1 8B, and Gemma 2 9B. Thanks to this technological advancement, the direct implementation of these models in devices leads to numerous advantages:

• Reduced latency
• Increased reliability
• Cost savings
• Energy efficiency

A boost for the distribution of Mistral AI

The models should soon become available on theAI Hub from Qualcomm, although no specific date has yet been given. Currently, developers can now explore other models of Mistral AI, including Mistral 7B v0.3. This partnership with Qualcomm represents a major step forward for the French start-up.

Reactions and perspectives

Arthur Mensch, co-founder and CEO of Mistral AI, expressed his enthusiasm about this collaboration: “This collaboration with Qualcomm is an important milestone for Mistral AI, showing our ability to run our new models locally on devices equipped with Snapdragon platforms. This allows for faster processing, thereby reducing costs and energy requirements.”

For several months, Mistral AI has been intensifying its partnership initiatives with major technological players. The objective is clear: to expand its distribution network for its innovative models. Recent partnerships include:

• In March, Snowflake invested in Mistral AI while offering to distribute OpenAI’s LLMs on its platform.
• Microsoft, in November 2023, integrated Mistral AI models into its Azure platform.
• Google also signed an agreement to make Mistral-7B available via Vertex AIas well as the marketing of Mistral-8x7B on Google Cloud Marketplace.
• AWS has revealed the availability of the Mistral 7B and Mixtral 8x7B models on Amazon Bedrock.

Conclusion

The partnership between Qualcomm and Mistral AI marks a significant milestone in the development and integration of artificial intelligence models into various devices. Thanks to this collaboration, technological innovation continues to grow, bringing new horizons for local data processing. The future looks bright for Mistral AI, which is establishing itself as a key player in the AI ​​ecosystem.