- Mistral AI is a French company selling artificial intelligence (AI) products. It was founded in April 2023 by previous employees of Meta Platforms and Google DeepMind.
- Two models, Mistral 7B and Mixtral 8x7B, have been published and are available as weights. Three models, Mistral Small, Mistral Medium and Mistral Large, are available via API only.
- Mistral AI has also launched a chatbot called Le Chat, a counterpoint to ChatGPT and contained a partnership with tech giant Microsoft.
Perplexity AI is a young company that specialises in AI and machine learning solutions. They focus on developing advanced algorithms and technologies to tackle complex problems across various industries, including finance, healthcare, and technology.
What is Mistral AI?
Mistral AI, a French company selling AI, was founded in April 2023 by previous employees of Meta Platforms and Google DeepMind. It produces open-source large language models (LLMs) citing the foundational importance of open-source software and as a response to proprietary models.
Before co-founding Mistral AI, Arthur Mensch worked at Google DeepMind which is Google’s artificial intelligence laboratory, while Guillaume Lample and Timothée Lacroix worked at Meta Platforms. The co-founders met while students at École Polytechnique.
Mistral’s open-weight models
Two models, Mistral 7B and Mixtral 8x7B, have been published and are available as weights.
Its first language processing model “Mistral 7B” was available on 27 September 2023 under the free Apache 2.0 license. This model has 7 billion parameters, a small size compared to its competitors.
The company released the Mixtral 8x7B model with 46.7 billion parameters but using only 12.9 billion per token thanks to the mixture of experts’ architecture, on December 11, 2023. The model masters 5 languages (French, Spanish, Italian, English and German) and outperforms, according to its developers’ tests, the “LLama 2 70B” model from Meta.
Also read: French AI startup Mistral shakes things up with surprise release of LLM that’s better than ChatGPT
Mistral‘s API-only models
Three models, Mistral Small, Mistral Medium and Mistral Large, are available via API only, which means these models are closed-source and only available through the Mistral Application Programming Interfaces.
Microsoft announced a new partnership with the company in February to expand its presence in the rapidly evolving AI industry. Under the agreement, Mistral’s rich language models will be available on Microsoft’s Azure cloud, while the multilingual conversational assistant “Le Chat” will be launched in the style of ChatGPT.
Among all the Large models currently accessible through the API, Mistral Large ranks second, right after the GPT-4, and is the only one to score more than 80 points on the MMLU exam.
With the launch of Mistral Large, Mistral AI has also launched a chatbot called Le Chat, a counterpoint to ChatGPT, to replicate OpenAI’s successful path. Even with the support of Microsoft Azure computing resources, Le Chat’s servers are still crowded.
Also read: France’s Mistral launches Le Chat to challenge ChatGPT
In terms of inference accuracy, Mistral Large has surpassed Claude 2, Gemini 1.0 Pro, GPT-3.5 and other well-known large models, and it also supports 32k token context Windows, supports precise instructions, and comes with function call capabilities. In reasoning speed, the Mistral Large surpasses even the GPT-4 and Google’s recently launched Gemini Pro.
Many open-source large model enthusiasts worry that Mistral AI will go from open to closed like OpenAI. According to the interview with the CEO of Mistral, not only will they continue to adhere to the open-source concept in the future, but at the same time, they will also introduce the most powerful closed-source model to compete in the business.
