Powered by Sentence Transformers
multi-qa-mpnet-base-dot-v1 is a Sentence Transformers model that encodes sentences and paragraphs into 768-dimensional embeddings optimized for semantic search using dot-product similarity. It is trained on large-scale question–answer pairs to retrieve relevant passages for user queries.
Output tokens per second · Higher is better
Seconds · Lower is better
USD per 1M tokens (blended) · Lower is better
About the model
multi-qa-mpnet-base-dot-v1 is a sentence-transformers text-embedding model based on mpnet-base that maps natural-language inputs to 768-dimensional dense vectors. It is mainly used for semantic search, where both queries and candidate passages are embedded and ranked via dot-product similarity. It is also applied to related tasks like text similarity, clustering, and information retrieval in downstream applications. The model is part of the Sentence Transformers family and is fine-tuned from the pretrained mpnet-base transformer encoder.
Model capabilities
Enables high-quality semantic search by encoding queries and documents into a shared vector space for similarity-based retrieval.
Optimized for multi-domain question-answer retrieval, matching user questions to the most relevant passages or FAQ-style answers.
Produces dense sentence and passage embeddings capturing semantic meaning, suitable for clustering, ranking, and downstream NLP tasks.
Supports multilingual question matching, allowing questions in different languages to be mapped into a common embedding space for retrieval.
Computes cosine similarity between query and document embeddings to score relevance for information retrieval and recommendation pipelines.
Use cases
Transparent pricing
LLM API embeddings for multi-qa-mpnet-base-dot-v1 are up to ~60% cheaper than major providers
| Provider | Region | Latency | Throughput | Uptime | Input ($/1M) | Output ($/1M) | Context |
|---|---|---|---|---|---|---|---|
| LLM API BEST | Global | ~120ms | ~9000 tokens/s | 99.99% | $0.02 per 1M tokens | $0.00 per 1M tokens | ~8K tokens |
| Sentence Transformers (Hosted API) | Global | ~250ms | ~3000 tokens/s | ~99.9% | ~$0.05 per 1M tokens | $0.00 per 1M tokens | ~8K tokens |
| Hugging Face Inference API | EU West | ~280ms | ~2500 tokens/s | ~99.9% | ~$0.06 per 1M tokens | $0.00 per 1M tokens | ~8K tokens |
| Azure AI (Custom Container) | US East | ~220ms | ~4000 tokens/s | 99.9% | ~$0.07 per 1M tokens | $0.00 per 1M tokens | ~8K tokens |
| Replicate | Global | ~300ms | ~2000 tokens/s | ~99.5% | ~$0.08 per 1M tokens | $0.00 per 1M tokens | ~8K tokens |
Performance benchmarks
| Metric | multi-qa-mpnet-base-dot-v1 | all-mpnet-base-v2 | msmarco-distilbert-base-v3 |
|---|---|---|---|
| Model Type | Text embedding (bi-encoder) | General-purpose text embedding | MS MARCO passage ranking bi-encoder |
| Dimensions | 768 | 768 | 768 |
| Max Input Tokens | ~512 | ~512 | ~512 |
| Price per 1M Tokens | ~$0.10 | ~$0.10 | ~$0.08 |
| Avg Latency per 1K Tokens | ~120ms | ~130ms | ~110ms |
| Throughput | ~1,500 tps | ~1,400 tps | ~1,600 tps |
| Uptime | ~99.9% | ~99.9% | ~99.9% |
30-day usage via LLM API
Architecture & Integration
One unified API. Every major model. Built-in reliability, cost control, and observability.
Dynamically route requests across providers based on latency, cost, or quality policies. One API abstracts model sprawl so you ship faster with less integration work.
One endpoint, every model.Automatically balance model choice and token usage to hit your budget targets. Set cost policies once and let LLM.API optimize every call for price-performance.
Lower spend, same output.Define per-route fallback chains so outages or rate limits never break production. LLM.API transparently retries on alternate models while preserving contracts and payloads.
Stay online, automatically.Trace every request across providers with unified logs, metrics, and evaluations. Debug latency, failures, and quality issues from a single, provider-agnostic dashboard.
See every token hop.Declare tasks like chat, embeddings, tools, or rerank once—LLM.API handles prompt shaping and provider quirks so your code stays clean and portable.
Code to tasks, not vendors.Send thousands of requests in a single batch with built-in concurrency control and retries. Maximize throughput while keeping provider limits and costs in check.
Scale workloads, not code.Decision guide
FAQ
multi-qa-mpnet-base-dot-v1 is a Sentence Transformers model for generating dense embeddings tailored to multi-domain question answering and semantic search.
It is best for semantic search, duplicate question detection, information retrieval, and building QA systems over documents or knowledge bases.
LLM.API usage-based pricing applies per embedding request; check the multi-qa-mpnet-base-dot-v1 pricing section in your LLM.API dashboard.
multi-qa-mpnet-base-dot-v1 is typically used with inputs up to a few hundred tokens; very long texts should be chunked before embedding.
On LLM.API, it usually returns embeddings in tens of milliseconds to low hundreds per request, depending on batch size and load.
multi-qa-mpnet-base-dot-v1 is a text-only model that takes natural language strings and outputs fixed-size vector embeddings.
Use the LLM.API embeddings endpoint, set the provider to Sentence Transformers, and specify multi-qa-mpnet-base-dot-v1 as the model name.
It generally offers strong retrieval quality with moderate embedding size and speed, often outperforming older MiniLM-based QA embeddings.
It works best for English; performance on other languages may be weaker and is not specifically optimized for broad multilingual coverage.
It cannot generate text, handle images, or reason over very long documents without chunking, and embedding quality degrades on noisy or non-English data.
Compare
Nemotron 3 Super is NVIDIA’s open-weight, 120B-parameter hybrid Mamba-Transformer Mixture-of-Experts language model optimized for high-throughput agentic reasoning workloads. It is notable for combining LatentMoE experts, long-context…
GPT-5.4 Mini is an OpenAI language model variant optimized for lightweight, general-purpose assistant tasks. It is designed to balance capability with efficiency for everyday conversational and…
GLM 4.6 is Z.ai’s flagship mixture-of-experts large language model optimized for coding, reasoning, and agentic workflows. It is notable for its strong performance on code benchmarks…
