Mistral AI has introduced its newest model, the Mistral Large 2, in response to Meta and OpenAI’s recently released models. Mistral Large 2 excels in code generation, mathematics and multilingual support. With 128 thousand context windows, it improves user experience in various applications by enabling the processing of comprehensive data inputs.
The launch of the Mistral Large 2 comes just one day after Meta released its latest and greatest open source model, the Llama 3.1 405b.
Mistral says the Large 2 raises the performance and cost bar for open models, and a handful of benchmarks back this up. The Large 2 appears to beat the Llama 3.1 405B in code generation and math performance, and does so with less than a third of the parameters.
Mistral specially trained Large 2 to notice things he doesn’t know
What sets the Mistral Large 2 apart is its focus on minimizing situations where it produces plausible but inaccurate information. Because Mistral specially trained Large 2 to notice things he does not know, which is actually a very important development.
of the model A window of 128 thousand tokens There is, which means Large 2 can receive a lot of data in a single request. Mistral’s new model also includes advanced multilingual support. Large 2, English, French, German, Spanish, Italian, Portuguese, Arabic, Hindi, Russian, Chinese, Japanese ve Koreanas well as 80 coding languages He also understands.
Mistral also claims that Large 2 produces more concise answers than leading AI models. Large 2 can be used on Google Vertex AI, Amazon Bedrock, Azure AI Studio and IBM watsonx.ai.
Source link: https://webrazzi.com/2024/07/25/mistral-ai-in-yeni-yapay-zeka-dil-modeli-large-2/