Zenaique

Compute the cost to embed 1B tokens with text-embedding-3-large via batch API

Predict output·Medium·4.0 · 0·~2 min·Asked atAi4bharatN8nPersistent·Relevant atElasticHugging FaceNeo4jQdrant
Attempt it
You're embedding 1 billion tokens of corpus with OpenAI text-embedding-3-large (sync price: $0.13 per 1M tokens) via the batch API, which gives a 50% discount. Compute the total cost in USD.
TL;DR

$65. 1B tokens at $0.13 per million is $130 sync; the batch API halves it to $65. Still cheap relative to storage and the cost of repeat re-embeddings.

Memory aid
Sign in to see the mnemonic that makes this stick.
Easy to grasp

Imagine buying concert tickets in bulk. At the box office window they cost thirteen cents each, but if you mail in your order and let them ship by tomorrow, the price drops to six and a half cents each. You need one thousand tickets. At the cheaper rate that comes to sixty-five dollars total. That is exactly the cost of converting a billion pieces of text into the premium catalog cards using the overnight lane. The bill looks tiny, almost laughable next to a real engineering budget. The expensive part shows up later, when the catalog system gets updated and every single card has to be reprinted from scratch in the new format. That reprint, repeated across years, is where the real money goes, not the original sixty-five dollars.

Key concepts

Concept explanation~2 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

This question is the sibling of the text-embedding-3-small version, with two extra pieces of arithmetic: the larger model's per-token price and the batch API discount. The interview value remains in the framing: what the dollar number tells you about the real cost structure of an embedding stack.

The candidate who computes $65 and stops has answered the literal question. The candidate who computes $65 and then walks through 'this is still tiny; the storage doubling is what actually compounds' has shown they understand production embedding economics.

The two-step arithmetic

Start with the sync price. text-embedding-3-large costs $0.13 per million tokens on the sync API in 2026. One billion tokens divided by a million per unit is 1,000 units. 1,000 * $0.13 = $130 sync.

Apply the batch discount. OpenAI's batch API charges half the sync price in exchange for an asynchronous completion model with a 24-hour SLA per submission. Half of $130 is $65.

The answer is $65. The sanity check: $0.065 per million times 1,000 millions is also $65, confirming the arithmetic.

Comparison points worth knowing. text-embedding-3-small at $0.01 batch on the same billion-token corpus is $10. Voyage-3 at $0.18 per million is $90 batch (Voyage offers a 50% batch discount). Cohere embed-v4 at $0.10 per million is $50 batch. The OpenAI text-embedding-3 family is competitive but not cheapest in 2026.

Why the larger model costs more and what you get
Where the cost compounds across the index lifetime
When sync vs batch makes sense and operational notes
Sign in to unlock the full deep dive.

Situations where this technique stops working.

Sign in to see when this approach fails.

2–4 min · Everything important, quickly.

Sign in to see the quick scan of the deep dive.

Real products, models, and research that use this idea.

  • OpenAI's 2026 pricing page lists text-embedding-3-large at $0.13 per million tokens sync and $0.065 batch.
  • Production RAG teams routinely migrate from text-embedding-3-small to text-embedding-3-large via parallel-index migration, paying the one-time batch bill in exchange for the recall lift.
Sign in to see more production examples.

What an interviewer would ask next. Try answering before peeking at the approach.

QWhy is text-embedding-3-large priced 6.5x higher than text-embedding-3-small?
A

Bigger encoder model, more FLOPs per token, larger output vector. The price differential approximates the compute differential at OpenAI's serving margins.

2 more follow-ups an interviewer would ask next. Sign in to reveal them.

Red flags & common mistakes

The phrases that signal junior thinking. Click to expand.

Most common mistake

Forgetting to apply the batch discount, or applying it to the wrong base price; the sync number is $130, batch is $65.

Sign in to see all red flags and common mistakes.

60 second bullets to scan on the way to the call.

  • Sync vs batch API pricing for text-embedding-3-large

  • 6.5x cost ratio vs text-embedding-3-small

Sign in to unlock the revision sheet.

Primary sources. Browse if you want the original framing.

Similar questions

Same topic, related formats. Practice these next.

4 curated
Next question
Why is cosine similarity preferred over Euclidean distance for text embeddings?
Flashcard·Easy