Mistral NeMo
The most advanced 12B model, supporting multilingual applications
- Provide a large contextual window of up to 128k tokens
- Supports multiple languages, including but not limited to English, French, German, Spanish, Italian, Portuguese, Chinese, Japanese, Korean, Arabic, and Hindi
- Using Tekken tokenizer to improve the compression efficiency of text and source code
- Through instruction fine-tuning, the ability to follow precise instructions, reason, handle multiple rounds of conversations, and generate code has been improved
- Perform FP8 inference without sacrificing performance
- The pre trained basic checkpoints and instruction adjusted checkpoints are both released under the Apache 2.0 license
- The model weights are hosted on HuggingFace and support both Mistral reference and Mistral finetune
Product Details
Mistral NeMo is a 12B model jointly built by Mistral AI and NVIDIA, with a large context window of 128k tokens. It is in a leading position in reasoning, world knowledge, and coding accuracy. This model is designed for global multilingual applications, supporting multiple languages such as English, French, German, Spanish, Italian, Portuguese, Chinese, Japanese, Korean, Arabic, and Hindi. Mistral NeMo also uses a new tokenizer Tekken, which improves the compression efficiency of text and source code. In addition, the model has been fine tuned with instructions to enhance its ability to follow precise instructions, reason, handle multiple rounds of dialogue, and generate code.
