- Text Embeddings
bge-m3 is a multilingual text embedding model from BAAI that produces dense, sparse, and ColBERT-style multi-vector embeddings in a single pass, enabling powerful hybrid retrieval. It…
Powered by OpenAI
Text Embedding Ada 002
text-embedding-ada-002 is an OpenAI embedding model that converts text into numerical vectors for measuring semantic similarity. It is an improved, more performant successor to earlier Ada-based embedding models and became a widely used default for production applications.
Providers Comparison
Speed
Output tokens per second · Higher is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
End-to-End Response Time
Seconds · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
Price
USD per 1M tokens (blended) · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
About the model
What is Text Embedding Ada 002?
text-embedding-ada-002 is an OpenAI text embedding model that maps text into high-dimensional vectors representing semantic meaning. It is mainly used for semantic search and retrieval, where both documents and queries are embedded and compared via vector similarity. It is also widely used for tasks like clustering, recommendations, anomaly detection, and lightweight classification built on top of embeddings. text-embedding-ada-002 is part of OpenAI’s Ada family of models and was introduced as a unified replacement for several older text and code search/similarity models.
Model capabilities
5 Core Capabilities
-
Text Embedding
Generates dense vector representations of text that capture semantic meaning for use in search, clustering, and classification tasks.
-
Semantic Search
Enables retrieval of relevant documents by comparing embedding vectors, supporting meaning-based search beyond simple keyword matching.
-
Text Clustering
Supports grouping of similar texts by comparing embeddings, enabling topic discovery, content organization, and deduplication workflows.
-
Multilingual Embeddings
Produces embeddings for multiple languages in a shared vector space, enabling cross-lingual similarity search and analysis.
-
Anomaly Detection
Helps identify outlier texts by comparing embedding distances, useful for spotting unusual content or potential data quality issues.
Use cases
6 Most Valuable Use Cases
- Semantic Text Search
- Document Clustering
- Recommendation Systems
- Anomaly Detection
- Topic Tagging
- Change Monitoring
Transparent pricing
Cost Comparison
LLM API offers the lowest cost per 1M embedding tokens with superior performance and reliability.
| Provider | Region | Latency | Throughput | Uptime | Input ($/1M) | Output ($/1M) | Context |
|---|---|---|---|---|---|---|---|
| LLM API BEST | Global | 80ms | 120K tps | 99.99% | $0.05 | $0.05 | 8192 tokens |
| OpenAI | Global | ~150ms | ~40K tps | 99.9% | $0.10 | $0.10 | 8192 tokens |
| Azure OpenAI | US East | ~170ms | ~35K tps | 99.9% | ~$0.11 | ~$0.11 | 8192 tokens |
| Anthropic-Compatible API | Global | ~160ms | ~30K tps | 99.9% | ~$0.12 | ~$0.12 | ~8000 tokens |
Performance benchmarks
Technical Specifications
| Metric | Text Embedding Ada 002 (OpenAI) | text-embedding-3-large (OpenAI) | text-embedding-004 (OpenAI) |
|---|---|---|---|
| Dimensions | 1536 | 3072 | 1536 |
| Max Input Tokens | 8K | 200K | 200K |
| Price per 1M Tokens (Input) | $0.10 | $0.13 | $0.02 |
| Avg Latency | ~120ms | ~150ms | ~140ms |
| Throughput | ~2,000 tps | ~1,800 tps | ~2,200 tps |
| Uptime | 99.9% | 99.9% | 99.9% |
30-day usage via LLM API
- 68B
- Embedding tokens processed (30 days)
- 24M
- Embedding API requests (30 days)
- 190K
- Active developer accounts (30 days)
- 99.98%
- Avg API uptime (30 days)
Architecture & Integration
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Intelligent Model Routing
Automatically send each request to the optimal model across providers based on cost, latency, and quality—without changing your integration or redeploying code.
One endpoint, any model -
Cost-Aware Orchestration
Dynamically balance premium and budget models using policy-based controls and real-time pricing, so you stay within budget while preserving the quality your app needs.
Control spend by design -
Resilient Fallback Logic
Handle provider outages and rate limits automatically with multi-provider fallbacks, ensuring your production workloads stay online even when individual APIs fail.
Always-on reliability -
End-to-End Observability
Trace every request across providers with logs, metrics, and structured events so you can debug prompts, tune routing, and prove reliability to stakeholders.
See every token -
Task-Level Abstractions
Use high-level tasks—chat, tools, RAG, function calling—instead of vendor-specific APIs, so you can swap models without rewriting business logic.
Code to tasks, not vendors -
High-Throughput Batch Jobs
Run massive batch inference across multiple providers with concurrency, retries, and progress tracking built in, turning offline workloads into a single API call.
Scale batch without ops
Decision guide
When to Use — When NOT to Use
Use it if...
- You need inexpensive, general-purpose text embeddings for search or recommendation systems.
- Your use case involves semantic search over large document collections with moderate accuracy needs.
- You need vector representations to power content-based recommendations, like similar articles or products.
- Your use case involves clustering or deduplicating short texts, titles, or product descriptions.
- You need multilingual embeddings for many languages without strict requirements for top-tier performance.
- Your use case involves intent or topic similarity matching in chatbots or support ticket routing.
- You need a simple, stable embedding model supported natively by the OpenAI platform.
Avoid if...
- You need state-of-the-art embedding performance on complex, nuanced semantic tasks across languages.
- Your workload requires embeddings specifically tuned for code understanding or repository search.
- You need domain-specialized embeddings, such as for legal, medical, or financial texts.
- Your workload requires cross-modal embeddings that jointly represent text and images or audio.
- You need fine-grained, sentence-level reasoning comparable to large modern language models directly.
- Your workload requires extremely long-context embeddings beyond typical input length constraints.
- You need strict data residency, on-prem, or self-hosted embedding solutions outside OpenAI.
FAQ
Frequently Asked Questions
-
What is Text Embedding Ada 002?
Text Embedding Ada 002 is an OpenAI model that converts text into numerical vector embeddings for tasks like semantic search, clustering, and recommendation.
-
What is Text Embedding Ada 002 best used for?
It is best for semantic similarity, search ranking, deduplication, recommendations, and representing documents or queries in a shared vector space.
-
What is the pricing for Text Embedding Ada 002 on LLM.API?
LLM.API forwards OpenAI’s token-based pricing for Text Embedding Ada 002; check the LLM.API pricing page for the latest per‑token rates.
-
What is the context window of Text Embedding Ada 002?
Text Embedding Ada 002 accepts up to roughly 8K tokens of input text per request, depending on exact tokenization.
-
How fast is Text Embedding Ada 002 in terms of latency?
Text Embedding Ada 002 is optimized for low-latency embedding generation, typically suitable for real-time or near real-time applications.
-
Which modalities does Text Embedding Ada 002 support?
Text Embedding Ada 002 supports only text input, producing numeric vector embeddings as output, and does not process images, audio, or video.
-
How do I call Text Embedding Ada 002 through LLM.API?
Use the LLM.API embeddings endpoint with the provider set to OpenAI and the model name set to text-embedding-ada-002.
-
How does Text Embedding Ada 002 compare to larger OpenAI embedding models?
Compared to larger models, Text Embedding Ada 002 typically offers lower cost and faster inference at slightly reduced embedding quality.
-
Are there any key limitations of Text Embedding Ada 002?
It cannot generate text, handle multimodal input, or understand context beyond its token limit; it only encodes text into fixed-size vectors.
-
Can I use Text Embedding Ada 002 for multilingual text?
Yes, it supports multiple languages, but embedding quality can vary across languages and may be strongest for English.
Compare
Competitive Models
- Text Embeddings
Gemini Embedding 2 is Google's natively multimodal embedding model that maps text, images, video, audio, and documents into a single semantic vector space. It is notable…
- Text Embeddings
E5-Base-v2 is an English sentence and paragraph embedding model from Intfloat that encodes text into a 768-dimensional dense vector space, optimized for high-quality semantic similarity tasks.