NVIDIA Mistral AI joined forces, and the king of 12 billion small models came on stage, crushing the Llama 3 single sheet with 4090 runs.
Mistral AI announced the release of Mistral NeMo, a 12B parametric model co-developed by NVIDIA with a context window of up to 128k tokens.
The model is designed to support enterprise applications, including chatbots, multilingual tasks, coding and summaries. In its size category, Mistral NeMo leads in reasoning, world knowledge and code accuracy. Using a standard architecture, Mistral NeMo is easy to use and serves as a direct replacement for any Mistral 7B system.
To encourage adoption and further research, Mistral AI has provided pre-training basics and instruction tuning checkpoints under the Apache 2.0 license. This open source approach may appeal to researchers and businesses, accelerating the integration of the model in various applications.
A key feature of Mistral NeMo is the awareness of quantification during the training process, which allows FP8 inference to be achieved without affecting performance. This capability can be critical for organizations that want to efficiently deploy large language models.
Mistral AI provides a performance comparison between the Mistral NeMo base model and two recent open source pre-trained models: Gemma 29B and Llama 38B.
Mistral NeMo introduced Tekken, a new marker based on Tiktoken. Tekken is trained in more than 100 languages and provides better compression efficiency of natural language text and source code than the SentencePiece tagger used in previous Mistral models. The company reports that Tekken has improved its efficiency by approximately 30% in compressing source code and several major languages, with the improvement being more significant for Korean and Arabic.
Mistral AI also claims that Tekken is superior to the Llama 3 tagger in terms of text compression and works in approximately 85% of all languages, which may give Mistral NeMo an advantage in multilingual applications.
Weights for the model are now available on HuggingFace, including base and instruction versions. Developers can start experimenting with Mistral NeMo using the mistral-inference tool and make adjustments through mistral-finetune. For users using the Mistral platform, this model is provided under the name open-mistral-nemo.
In tribute to its collaboration with NVIDIA, Mistral NeMo is also packaged as an NVIDIA NIM inference microservice and made available through ai.nvidia.com This integration may simplify deployments for organizations that are already investing in the NVIDIA AI ecosystem.
The release of Mistral NeMo represents an important advance in the democratization of advanced AI models. By combining high performance, multilingual capabilities and open source availability, Mistral AI and NVIDIA are positioning the model as a multifunctional tool widely used in various industries and research areas.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Official introduction:https://mistral.ai/news/mistral-nemo
Oil tubing: