title: "Mistral NeMo | Mistral AI"

Mistral AI has released Mistral NeMo, a 12B parameter model developed in collaboration with NVIDIA. The model features a large 128k token context window and achieves state-of-the-art performance in its size class, outperforming models like Gemma 2 9B and Llama 3 8B. Designed for global multilingual applications, it utilizes a new, more efficient tokenizer called Tekken, which offers significant compression improvements for various languages and source code. The pre-trained base and instruction-tuned checkpoints are available under the Apache 2.0 license on HuggingFace.