Deci AI

About Us

Deci is on a mission to empower AI teams with advanced tools to remove development barriers and attain production-grade performance and efficient inference faster.

If you have questions, please feel free to contact us through our website,
reach out to us on Twitter, or join our Deep Learning Daily Discord community.

DeciLM-7B:

A 7.04 billion-parameter decoder-only text generation model, licensed under Apache 2.0. DeciLM-7B is not only the most accurate 7B base model to date, but it also currently outpaces all models in its class with a throughput that is up to 4.4x that of Mistral-7B's. DeciLM-7B’s architecture is the result of Deci's Neural Architecture Search technology. The model was fine-tuned using LoRA on the SlimOrca dataset, creating DeciLM-7B-instruct.

DeciCoder 1B:

A permissively licensed 1.1 billion-parameter code generation model generated by Deci's Neural Architecture Search technology.
Equipped with a 2048-context window, DeciCoder 1B delivers a 3.5x increase in throughput, improved accuracy on the HumanEval benchmark, and reduced memory usage compared to widely-used code generation LLMs such as SantaCoder.

DeciLM 6B:

A permissively licensed, 5.7 billion-parameter pretrained text generation model using variable Grouped Query Attention (GQA) to achieve an optimal balance between performance and computational efficiency. Generated by Deci's proprietary Neural Architecture Search technology, AutoNAC™, DeciLM 6B delivers 15x the throughput of Llama 2 7B while maintaining comparable quality. DeciLM-6B was fine-tuned using LoRA for instruction-following on a subset of the OpenOrca dataset, creating DeciLM 6B-Instruct.

DeciDiffusion:

A permissively licensed, text-to-image latent diffusion model generated by Deci's Neural Architecture Search technology.
DeciDiffusion generates Stable Diffusion-caliber images 3x faster.

Infery-LLM:

The most advanced inference SDK for LLM optimization and deployment, Infery-LLM includes unique features such as optimized kernels, continuous batching, advanced selective quantization, ultra-efficient beam search, parallel execution, and more. To explore the full capabilities of Infery-LLM, we invite you to try it out here.

YOLO-NAS:

An object detection foundational model generated by Deci's Neural Architecture Search technology. YOLO-NAS is a game-changer in the world of object detection, providing superior real-time object detection capabilities and production-ready performance.

SuperGradients:

An open-source library for training PyTorch-based computer vision models. Developed by Deci's deep learning experts for the benefit of the AI community, SuperGradients boosts performance with advanced techniques and enables users to easily train or fine-tune SOTA computer vision models for all main tasks with one training library.

DataGradients:

DataGradients is an open-source, Python-based library specifically designed for computer vision dataset analysis. It automatically extracts features from your datasets and combines them all into a single user-friendly report.

Deci Deep Learning Platform:

Simplify and accelerate the development of computer vision, NLP, and Generative AI applications with highly accurate and efficient foundation models. Utilize advanced tools to customize, optimize, and deploy deep learning models to production.

Deci is powered by a groundbreaking Automated Neural Architecture Construction (AutoNAC™) technology. Deci's AutoNAC™ engine democratizes the use of Neural Architecture Search for every organization and helps teams quickly generate fast, accurate, and efficient deep learning models.