Loading page
About this tool
Calculate the memory an embedding vector needs from its dimension count and precision, with per-1K and per-1M totals.
This calculator computes the exact memory one embedding vector occupies — dimensions multiplied by bytes per dimension, so a 1536-dimension float32 embedding takes 6,144 bytes. It is built for engineers sizing a RAG pipeline or vector database who need per-vector, per-thousand and per-million figures before choosing a model or a quantisation scheme, across float64, float32, float16, bfloat16, int8 and binary storage.
Open Embedding Dimension Calculator on AltFTool — it loads instantly in your browser.
Enter the values you already know.
Fine-tune the options to match your scenario.
Read the result and use it in your planning or reporting.
Vector size is deterministic arithmetic: ceil(dimensions × bits per dim ÷ 8), reported per vector and at 1K, 1M and 10M scale.
One table shows the same vector in float64 down to 1-bit binary, with the percentage saved versus float32.
One tap loads the published dimension counts of OpenAI, Cohere, BGE, E5 and MiniLM embedding models.
6,144 bytes (6 KiB) in float32, because each of the 1,536 dimensions takes 4 bytes. The same vector is 3,072 bytes in float16, 1,536 bytes in int8 and 192 bytes with binary quantisation. A million such float32 vectors need about 5.7 GiB before any index overhead.
float32 (4 bytes per dimension) is the default in Faiss, pgvector, Milvus, Qdrant and most managed services. Many of them offer scalar (int8) or binary quantisation as an opt-in to cut memory by 4x to 32x, usually at a small recall cost that a reranking step can recover.
Multiply the dimension count by the bytes per dimension of the storage format: dimensions × 8 for float64, × 4 for float32, × 2 for float16 or bfloat16, × 1 for int8, and ÷ 8 for binary (one bit per dimension). Add whatever metadata — ids, source text, JSON — you store alongside each vector.
Yes — storage scales exactly linearly with dimensions, so truncating OpenAI text-embedding-3-large from 3072 to 1024 dimensions cuts vector storage by two-thirds. Models trained with Matryoshka representation learning are designed so shortened vectors keep most of their retrieval quality.
Add the Embedding Dimension Calculatorwidget to your blog or website — free, responsive, no signup. Just keep the “Widget by AltFTool” credit link visible.
<iframe src="https://www.altftool.com/embed/widget/embedding-dimension-calculator"
title="Embedding Dimension Calculator — free AltFTool widget"
width="100%" height="640" style="border:0;border-radius:12px;overflow:hidden"
loading="lazy" referrerpolicy="no-referrer-when-downgrade"></iframe>
<p style="font-size:12px;margin:4px 0 0">Widget by <a href="https://www.altftool.com/tools/all/embedding-dimension-calculator?utm_source=embed&utm_medium=widget">AltFTool — free online tools</a></p>