Mistral AI CEO announces open source GPT-4 level model in 2024
Arthur Mensch, CEO of French startup Mistral AI, announced on French national radio that the company plans to open source the GPT-4 level model in 2024, which attracted attention in the industry.
Recently, Mistral AI launched the 8X7B model based on MoE architecture, which is comparable to other popular models such as GPT3.5 and Llama270B, and surpasses Llama270B with 6x faster inference speed in most benchmarks. This model is licensed under the Apache 2.0 license, providing developers with greater flexibility.
After the news spread, it caused a heated discussion on social media. One user expressed concerns about Mistral AI's future profitability, especially after the release of the 7B model they call "Mistral Tiny." This also creates curiosity about the upcoming “Mistral Medium” model.
Mistral AI divides its models into three categories: Mistral Tiny, Mistral Small, and Mistral Medium. The currently released Mistral7B Instruct v0.2 and Mixtral8x7B belong to Mistral Tiny and Mistral Small respectively. Interestingly, the Mistral Medium model hasn't been released yet, but the company says it's in development and will top the charts on standard benchmarks.
The company recently announced it has received $415 million in financing, valuing it at $2 billion. The latest round of financing was led by Andreessen Horowitz (a16z), and Lightspeed Venture Partners also joined as a further investment. In response to the commercial sustainability issues faced by open source big language model companies, Mistral AI recently launched "La Plateforme" to provide API endpoints for its available models.
In addition, Mistral AI is working on developing a Mistral Medium model that supports multi-language, which performs well in English, French, Italian, German, Spanish and code, reaching 8.6 points in the MT-Bench score, even surpassing GPT3 .5.
It is worth noting that at the same time, rumors that OpenAI may release GPT-4.5 at the end of the year are also spreading in the industry. This makes 2024 full of expectations and competition for the field of open source large language models.