text-embedding-3-small is OpenAI's cost-effective embedding model, released in January 2024. It converts text into 1536-dimensional vectors that capture semantic meaning — the foundation for semantic search, RAG pipelines, recommendation systems, and classification tasks.
This guide covers everything: pricing, token limits, dimensions, API usage, dimension reduction, performance benchmarks, and how it compares to text-embedding-3-large.
| Spec | Value |
|---|---|
| Model name | text-embedding-3-small |
| Provider | OpenAI |
| Default dimensions | 1536 |
| Adjustable dimensions | 256 – 1536 |
| Max input tokens | 8,191 |
| Max batch size | 2,048 inputs per request |
| Pricing (OpenAI direct) | $0.020 per 1M tokens |
| Pricing (Crazyrouter) | $0.016 per 1M tokens |
| MTEB benchmark score | 62.3 |
| Multilingual | Yes |
| Output format | float or base64 |
| Release date | January 25, 2024 |
| Status | Active (not deprecated) |
text-embedding-3-small costs 0.020per1milliontokens∗∗onOpenAIdirectly.Through[Crazyrouter](https://crazyrouter.com),thepricedropsto∗∗0.020 per 1 million tokens** on OpenAI directly. Through [Crazyrouter](https://crazyrouter.com), the price drops to **0.016 per 1M tokens — a 20% discount.
To put that in perspective:
| Document Volume | Approx. Tokens | Cost (OpenAI) | Cost (Crazyrouter) |
|---|---|---|---|
| 100 pages of text | ~75,000 | $0.0015 | $0.0012 |
| 10,000 pages | ~7.5M | $0.15 | $0.12 |
| 1 million pages | ~750M | $15.00 | $12.00 |
| Wikipedia (English, full) | ~4.4B | $88.00 | $70.40 |
| Model | Price / 1M Tokens | Price via Crazyrouter | Dimensions |
|---|---|---|---|
| text-embedding-3-small | $0.020 | $0.016 | 1536 |
| text-embedding-3-large | $0.130 | $0.100 | 3072 |
| text-embedding-ada-002 | $0.100 | — | 1536 |
| Google text-embedding-005 | $0.00625 | $0.005 | 768 |
| Cohere embed-v4 | $0.100 | — | 1024 |
| Voyage voyage-3-large | $0.180 | — | 2048 |
text-embedding-3-small is 6.5x cheaper than text-embedding-3-large and 5x cheaper than the older ada-002 — while outperforming ada-002 on benchmarks.
For a deeper comparison of all embedding models, see our AI Embeddings Comparison 2026 Guide.
The default output is a 1536-dimensional vector. But text-embedding-3-small supports dimension reduction via the dimensions parameter — you can request any value from 256 to 1536.
This is done using Matryoshka Representation Learning (MRL). The model is trained so that the first N dimensions of the vector carry the most important information. Truncating to fewer dimensions loses some nuance but keeps most of the semantic signal.
| Dimensions | MTEB Score | Storage per Vector | Relative Quality |
|---|---|---|---|
| 1536 (default) | 62.3 | 6,144 bytes | 100% |
| 1024 | ~61.5 | 4,096 bytes | ~98.7% |
| 768 | ~60.8 | 3,072 bytes | ~97.6% |
| 512 | ~59.7 | 2,048 bytes | ~95.8% |
| 256 | ~57.8 | 1,024 bytes | ~92.8% |
text-embedding-3-small accepts up to 8,191 tokens per input string. This is the model's context window for embedding.
Key details:
cl100k_base (same as GPT-4)For documents longer than 8,191 tokens, you have two options:
Send up to 2,048 texts in a single request for better throughput:
For bandwidth-sensitive applications, request base64-encoded output:
This is the most common comparison. Here's the full breakdown:
| Feature | text-embedding-3-small | text-embedding-3-large |
|---|---|---|
| Default dimensions | 1536 | 3072 |
| Adjustable dimensions | 256 – 1536 | 256 – 3072 |
| Max tokens | 8,191 | 8,191 |
| MTEB score | 62.3 | 64.6 |
| Price / 1M tokens | $0.020 | $0.130 |
| Price via Crazyrouter | $0.016 | $0.100 |
| Relative quality | Baseline | +3.7% better |
| Relative cost | Baseline | 6.5x more expensive |
| Storage (float32) | 6 KB/vector | 12 KB/vector |
text-embedding-3-large at 1024 dimensions often outperforms text-embedding-3-small at 1536 dimensions — while using less storage:
text-embedding-ada-002 was OpenAI's previous generation embedding model. Here's why you should migrate:
| Feature | text-embedding-3-small | text-embedding-ada-002 |
|---|---|---|
| MTEB score | 62.3 | 61.0 |
| Price / 1M tokens | $0.020 | $0.100 |
| Dimensions | 1536 (adjustable) | 1536 (fixed) |
| Dimension reduction | ✅ Yes | ❌ No |
| Multilingual (MIRACL) | 44.0 | 31.4 |
text-embedding-3-small is 5x cheaper, higher quality, and supports dimension reduction. There's no reason to stay on ada-002.
Migration note: Embeddings from different models are not compatible. Switching requires re-embedding all your documents.
text-embedding-3-small supports multilingual text natively. On the MIRACL multilingual benchmark, it scores 44.0 — a massive improvement over ada-002's 31.4.
Supported languages include English, Chinese, Japanese, Korean, Spanish, French, German, Portuguese, Russian, Arabic, Hindi, and many more.
For applications that are primarily multilingual (100+ languages, cross-lingual retrieval at scale), also consider Cohere embed-v4 or the open-source BGE-M3.
No. As of 2026, text-embedding-3-small is fully active and supported by OpenAI. It is not deprecated and there is no announced deprecation date.
The model that is deprecated is the older text-embedding-ada-002. OpenAI recommends migrating from ada-002 to the text-embedding-3 series.
Timeline:
Here's a complete working example:
For production workloads with millions of documents, use a vector database instead of in-memory numpy:
Other popular vector databases that work well with text-embedding-3-small:
| Model | Average | Classification | Clustering | Retrieval |
|---|---|---|---|---|
| text-embedding-3-small | 62.3 | 67.1 | 41.2 | 51.7 |
| text-embedding-3-large | 64.6 | 69.8 | 44.1 | 55.4 |
| text-embedding-ada-002 | 61.0 | 66.3 | 40.1 | 49.3 |
| Cohere embed-v4 | 66.2 | 71.5 | 46.3 | 56.8 |
| Voyage voyage-3-large | 67.1 | 72.0 | 47.2 | 58.1 |
| Model | MIRACL Score |
|---|---|
| text-embedding-3-small | 44.0 |
| text-embedding-3-large | 54.9 |
| text-embedding-ada-002 | 31.4 |
text-embedding-3-small is one of the fastest commercial embedding models. Typical latency through Crazyrouter:
| Batch Size | Avg Latency |
|---|---|
| 1 text | ~50ms |
| 10 texts | ~80ms |
| 100 texts | ~200ms |
| 1000 texts | ~800ms |
usage.total_tokens in responses to manage coststext-embedding-3-small is OpenAI's cost-effective text embedding model. It converts text into 1536-dimensional numerical vectors that capture semantic meaning, enabling applications like semantic search, RAG, classification, and clustering.
0.020per1milliontokensonOpenAIdirectly.Through[Crazyrouter](https://crazyrouter.com),itcosts0.020 per 1 million tokens on OpenAI directly. Through [Crazyrouter](https://crazyrouter.com), it costs 0.016 per 1M tokens — 20% cheaper. Embedding 10,000 pages of text costs roughly $0.15.
8,191 tokens per input string. That's approximately 6,100 words or 24,000 characters in English. Text exceeding this limit is silently truncated.
1,536 dimensions by default. You can reduce this to any value between 256 and 1,536 using the dimensions parameter in the API request.
No. It is fully active and supported as of 2026. The deprecated model is the older text-embedding-ada-002.
text-embedding-3-large outputs 3,072 dimensions (vs 1,536), scores 3.7% higher on MTEB benchmarks, and costs 6.5x more (0.13vs0.13 vs 0.02 per 1M tokens). For most applications, the small model is sufficient.
Yes. It handles multiple languages natively and scores 44.0 on the MIRACL multilingual benchmark. No special configuration is needed — just pass text in any supported language.
Yes. LangChain has built-in support:
Yes:
Open-source models like BGE-M3 (MTEB 63.2) and E5-Mistral-7B (MTEB 66.6) can match or exceed its quality. The trade-off is you need GPU infrastructure to run them. text-embedding-3-small wins on convenience and total cost for small-to-medium workloads.
text-embedding-3-small is the default choice for production embedding workloads in 2026. At 0.02/1Mtokens(or0.02/1M tokens (or 0.016 via Crazyrouter), it delivers strong quality across search, RAG, and classification — with the flexibility of dimension reduction and multilingual support.
Choose text-embedding-3-small when: you want the best price-to-performance ratio for embedding workloads.
Choose text-embedding-3-large when: you need maximum quality and the 6.5x cost increase is acceptable.
Access it through Crazyrouter for 20% lower pricing and a unified API that also gives you GPT-5, Claude, Gemini, and 300+ other models — all with one API key.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。