Loading…
Embedding costs across models: dimensions, $/1M tokens, workload totals.
| Model | Dims | Context | $/1M tokens | Per run | Per month |
|---|---|---|---|---|---|
| Gemini text-embedding-004cheapest | 768 | 2,048 | $0 | $0.0e+0 | $0 |
| OpenAI text-embedding-3-small | 1,536 | 8,191 | $0.02 | $0.01 | $0.01 |
| Cohere embed-v4 (int8) | 1,024 | 128,000 | $0.035 | $0.0175 | $0.02 |
| Voyage voyage-3 | 1,024 | 32,000 | $0.06 | $0.03 | $0.03 |
| OpenAI text-embedding-ada-002 | 1,536 | 8,191 | $0.1 | $0.05 | $0.05 |
| Cohere embed-v4 | 1,536 | 128,000 | $0.11 | $0.055 | $0.06 |
| OpenAI text-embedding-3-large | 3,072 | 8,191 | $0.13 | $0.065 | $0.07 |
🧮 Prices checked 2026-08; embedding input is cheap enough that most workloads cost cents — storage and retrieval usually dominate real RAG bills. Dimensions affect vector-DB storage more than API cost. Gemini text-embedding-004 is free-tier within quotas.
At $0.02 per million tokens, embedding a 100k-document corpus costs about a dollar once. Recompute frequency and document size — not API price — drive real bills, which the monthly column makes visible.
3072-dim vectors cost 3x the vector-DB storage of 1024-dim ones. The dims column exists so you can weigh retrieval quality against infrastructure costs, not just API price.
text-embedding-3-small wins most retrieval tasks at 6.5x lower price than large; upgrade only when benchmarked recall on your data actually improves. Int8 quantized Cohere halves dims with minor quality loss.
Typical RAG spends more on generation tokens than embedding tokens by orders of magnitude. Optimize chunk sizes and retrieval counts before agonizing over embedding model choice.
The Embedding Price Calculator lets you figure out embedding price calculatorinstantly, without reaching for a spreadsheet or doing the math by hand. Whether you're planning a budget, checking a loan, or working through homework, the tool applies the correct formula behind the scenes and shows the result the moment you enter your numbers.
Unlike a static chart or table, this calculator adapts to your exact inputs. You can adjust any value and see the outcome update in real time, which makes it easy to compare scenarios — for example, "what if the rate were 1% lower?" or "what if I paid an extra $50 a month?"
Common uses: people reach for this tool when they need to find a how much does openai embedding api cost, text-embedding-3-small vs large price, embedding cost per million tokens, or cheapest embedding api comparison.
Browser-based tools like this one have a few real advantages over installed software or manual methods:
The Embedding Price Calculator is based on the following formula:
Indexing cost = documents × tokens/doc ÷ 1,000,000 × P Monthly query cost = queries/day × tokens/query × 30 ÷ 1,000,000 × P Total first month = Indexing cost + Monthly query cost
Variables: documents = Number of texts to embed once (count) tokens/doc = Average tokens per document (tokens) queries/day = Search or retrieval queries per day tokens/query = Average tokens per query (tokens) P = Model price ($ per 1M tokens) dimensions = Embedding vector size (e.g. 1536); affects storage, not token price Indexing cost = One-time cost to embed the whole corpus ($) Monthly query cost = Recurring cost to embed daily queries ($)
Embeddings are billed only on input tokens, so cost is simply tokens ÷ 1,000,000 × price with no output component. The one-time corpus indexing is usually small next to recurring query traffic, and the price per million tokens is the main lever that separates providers on the same workload.
Worked example: Step 1: corpus = 50,000 documents × 400 tokens = 20,000,000 tokens. Step 2: Model A at $0.02 per 1M tokens → indexing = 20 × 0.02 = $0.40 one time. Step 3: traffic = 5,000 queries/day × 40 tokens = 200,000 tokens/day → × 30 = 6,000,000 tokens/month → 6 × 0.02 = $0.12/month. Step 4: Model B at $0.10 per 1M tokens → indexing = 20 × 0.10 = $2.00; queries = 6 × 0.10 = $0.60/month. Result: first-month total is $0.52 on Model A vs $2.60 on Model B — a 5× price gap passes straight through to the bill.
More tools you might find useful