The French Mistral AI startup has just unveiled a new open source model. According to the company, the model has better and faster performance with a low parameter than similar models of Google and Openai. Below you will read the specifications of this model of artificial intelligence.
According to the MISTRAL announcement, the new model that Mistral Small 3.1 The name processes the text and the image with only 24 billion parameters – the size of the current leading models – however, offers a performance similar to competing models or in some benchmarks.
New MISTRAL SPECIFIC SPECIFIC SPECIAL
“This new model comes with improved text performance, multifaceted understanding and a wide -field window up to 128,000 token,” says Mistral. The company claims that this model processs information at 150 token per second and is suitable for applications that require quick response time.
The company says the Mistral Small 3.1 works better than the GEMMA 3 Google and the Gpt-4O mini and is much faster:
With the release of this model under Apache 2.0, Mistral is pursuing a completely different strategy than its larger competitors, which have limited access to their strongest artificial intelligence systems. The new model is now available on the Hugging Face platform.
Mitsral says Mistral Small 3.1 is a lightweight model and you can run it on the RTX 4090 or Mac with 32 GB of RAM. Overall, this model is suitable for virtual assistants and other applications that require quick and accurate response.
In 2023, former researchers founded Google Dipmind and Mistral AI, quickly became a leading startup of artificial intelligence in Europe valuable about $ 6 billion. The company’s chats app, Le Chat, recently reached one million downloads in two weeks after its release. French President Emmanuel Macron also urged citizens to download citizens instead of ChatGpt or other chats.
Latest Passing over countries : Spain | Dominica | United Arab Emirates
RCO NEWS
