Machine Learning & AI

Mistral AI and CoreWeave Demonstrate Partnership at NVIDIA GTC, Mistral AI Hackathon

Mistral AI and CoreWeave Demonstrate Partnership at NVIDIA GTC, Mistral AI Hackathon

Mistral AI is currently the beacon of the open-source AI movement, and its latest release, Mistral Large, rivals Open AI’s ChatGPT-4 in performance.

There’s a new hero in the open-source community who is making a name for itself among the top-tier AI companies in the world. 

Meet Mistral AI, a ten-month-old  Paris-based startup that proved it can rank among the top in generative AI at a record-breaking pace. Over the last few months, Mistral AI has released world-class generative models, including Mistral-7B and Mixtral-8x7B, La Plateforme (their API service), and Le Chat (a "multilingual by design" chatbot).

Its latest and most advanced release, Mistral Large, is a new top-tier commercial model. Mistral AI’s performance on the suite of commonly used benchmarks makes it the world’s second-ranking model available through an API alongside GPT-4. This new flagship model has unique reasoning capacities, is proficient in code and mathematics, and is fluent in five languages (French, English, German, Spanish, and Italian—in line with Mistral AI’s commitment to promoting cultural and linguistic specificities in generative AI technology). It can process hundreds of pages of documents in a single call.

Figure 1: Comparison of GPT-4, Mistral Large (pre-trained), Claude 2, Gemini Pro 1.0, GPT 3.5, and LLaMA 2 70B on MMLU (Measuring massive multitask language understanding). Sourced from Mistral AI.

A Champion of the Open-Source Community

While the company is still quite new, Mistral AI’s leadership team is not. Founder and CEO Arthur Mensch left Google DeepMind to start Mistral AI with his former classmates, Timothée Lacroix and Guillaume Lample, who worked at Meta’s Paris AI lab.

There, they envisioned a radically new AI lab, one with a mission to make frontier AI ubiquitous and tailor-made with a commitment to sharing open-weight solutions. 

"Our ambition is to put frontier models in the hands of real-world application makers. Open source is a superb instrument to accelerate AI adoption."

— Arthur Mensch, founder and CEO at Mistral AI

Unlike other industry leaders that have prioritized closed-source models, Mistral AI has released both types: open and commercial  AI models. 

With its open models (Mistral-7B and Mixtral 8x7B) and optimized commercial models (Mistral Small, Mistral Medium, and Mistral Large), Mistral AI has become a standard-bearer for the open-source AI movement, developing models that out-perform many other LLMs from long-established and much larger companies.  

Part of the Mistral team in Paris, France. Sourced from Mistral AI.

Mistral AI and CoreWeave: Unrivaled Speed and Efficiency

All great race cars need strong and reliable engines. Mistral AI’s partnership with CoreWeave ensures they have the fastest and most flexible infrastructure for AI on the planet. 

Speed and efficiency have set Mistral AI apart from the early days of other AI labs. Just a few months after starting their company, the team released Mistral-7B last September, thanks in part to quick access to its cluster of NVIDIA H100 Tensor Core GPUs on CoreWeave Cloud. 

"We worked hand-in-hand with members of CoreWeave’s team to build the training clusters for our AI models—like having an extension of our own team on the infrastructure side. The performance we’ve seen from our CoreWeave clusters has been great, especially in terms of reliability and stability, and it’s enabled us to continue to deliver cutting-edge models quickly."

— Arthur Mensch, founder and CEO at Mistral AI

Mistral AI’s cost-efficiency is a testament to its masterful approach to model training and ultra-performant cluster on CoreWeave Cloud.

"Everything in AI moves fast, but Mistral AI moves faster. They’ve become an industry leader in a short period of time, and we’re proud to enable their acceleration."

— Max Hjelm, SVP of Revenue at CoreWeave

With ample funding, strong partnerships, fresh ideas, and now the world’s attention, Mistral AI is ready to steer the path forward for more open-source LLM achievements in generative AI.

See Mistral at NVIDIA GTC and Its Hackathon

At NVIDIA GTC, a global AI conference, Mistral AI’s CEO Arthur Mensch will sit down with CoreWeave’s SVP of Revenue, Max Hjelm, for a can’t-miss live conversation at booth 1208 at 2:00 p.m. Tuesday, March 19.

Arthur is also leading a session at 2:00 p.m. on Wednesday, March 20, at NVIDIA GTC, discussing Mistral AI’s fast achievements in open-source LLM model training and the scientific lessons learned while training its first models (Mistral and Mixtral).

After the event, Mistral AI is staying in the Bay Area over the weekend to host its first Hackathon, hosted on CoreWeave Cloud. 

See more about what CoreWeave is doing at NVIDIA GTC, and learn more about Mistral AI on their webpage.

Connect with us

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.