# FP8 Mixed-Precision Training **Source:** https://www.startuphub.ai/startups/fp8-mixed-precision-training **Profile type:** Startup > FP8 Mixed-Precision Training is a framework that uses 8-bit floating point operations to boost training throughput and reduce memory requirements for large Transformer models. ## About FP8 Mixed-Precision Training is a framework that uses 8-bit floating point operations to boost training throughput and reduce memory requirements for large Transformer models. It aims to achieve comparable accuracy to BF16 standards and enables efficient post-training quantization with up to 36% throughput gain. This approach is challenging due to potential numerical instability but is being addressed by frameworks that systematically suppress activation outliers. ## Key facts - **Slug:** fp8-mixed-precision-training - **Operating status:** Active - **Sectors / tags:** Foundation Model, Large Language Model, AI Infrastructure, AI Hardware, MLOps, Developer Tools, API Platform ## Frequently asked ### What does FP8 Mixed-Precision Training do? FP8 Mixed-Precision Training is a framework that uses 8-bit floating point operations to boost training throughput and reduce memory requirements for large Transformer models. It aims to achieve comparable accuracy to BF16 standards and enables efficient post-training quantization with up to 36% throughput gain. This approach is challenging due to potential numerical instability but is being addressed by frameworks that systematically suppress activation outliers. ### What industry does FP8 Mixed-Precision Training operate in? FP8 Mixed-Precision Training operates in Foundation Model, Large Language Model, AI Infrastructure, AI Hardware, MLOps, Developer Tools. ## JSON-LD Structured data aligned with this page's `