HomeIndustriesMistral AI releases recent model and chatbot for GPT-4

Mistral AI releases recent model and chatbot for GPT-4

French startup Mistral AI has released a chatbot and its flagship LLM (Large Language Model) called Mistral Large, which is pretty near the performance of GPT-4.

Mistral AI's meteoric rise made headlines in December when the corporate reached a $2 billion valuation lower than six months after its founding. Its lightweight, open-source Mixtral 8x7B LLM caused some excitement by beating larger competitors like Meta's Llama 2 34B.

There aren’t any details about what number of parameters Mistral Large has, but benchmark numbers released by Mistral AI show that its recent model outperforms Claude 2, Gemini Pro and Llama 2-70B. Its 32KB context window is smaller than a few of these models, but still equals about 20,000 words.

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 comprehension). Source: Mistral AI

OpenAI's GPT-4 stays the gold standard for comparing LLM performance, and Mistral Large comes pretty near it in various benchmarks.

Benchmark comparison of the Mistral Large with other leading models. Source: Mistral AI

The pricing model for API access to Mistral Large is one other foray from OpenAI. Mistral Large's query costs $8 per million input tokens and $24 per million output tokens. This is 20% cheaper than what OpenAI charges to question GPT-4.

While Mistral AI is a giant proponent of open source models, unlike Mixtral 8x7B, its recent models have only been released with API access.


Mistral AI also announced a beta version of its chatbot called Le Chat. The conversational chatbot allows users to experience Mistral Large responses, but additionally provides access to Mistral AI's recent smaller models, Mistral Small and Mistral Next.

The latter two LLMs offer more concise and favorable interactions. Mistral Small outperforms Mixtral 8x7B and has lower latency.

Mistral AI says it has tried to make its models “as useful and as least opinionated as possible.” The company says that with its “tunable, system-level moderation mechanism, le Chat will warn you in a non-invasive way when you find yourself pushing the conversation in directions where the assistant may produce sensitive or controversial content.”

So expect a little bit less alertness than twins might produce. You can try Le Chat Here.

Microsoft suggestion

Although Microsoft is OpenAI's largest backer, the corporate has made several other models available on its platforms. Coinciding with the discharge of Mistral Large, Microsoft announced that it might enter right into a multi-year partnership with Mistral AI and supply Mistral AI with its Azure platform to develop and deploy its models.

Arthur Mensch, CEO of Mistral AI, said: “We are more than happy about this partnership with Microsoft. With Azure’s state-of-the-art AI infrastructure, we’re reaching a brand new milestone in our expansion and bringing our revolutionary research and practical applications to recent customers world wide.”

With the discharge of the brand new Gemini models and Mistral Large following GPT-4, one has to wonder if the discharge of GPT-5 is imminent. If Mistral AI adds multimodal functionality to Mistral Large, pricing poses a serious threat to OpenAI.


Please enter your comment!
Please enter your name here

Must Read