Tech & Energy
Embedding Cost Calculator
Embeddings have gotten cheap. Vectorize a 10M-token corpus for under $2 with the right model.
Total cost
—
- Per 1M tokens
- —
- Vector dim
- —
- Storage (MB) @ float32
- —
FAQ
Smaller dims = worse retrieval? +
Surprisingly little. text-embedding-3-large 256-dim retains 90%+ of full-dim accuracy at 1/12 the cost. MRL is huge.
Local vs API? +
BGE-large on a single GPU embeds at ~1M tokens/min. Free after the GPU. Worth it if you have one already running.