Methodology
The selector uses documented embedding categories, supported tasks, or capabilities, then ranks by context fit, deployment match, provider breadth, and known input price.
Model Selection
Embedding Model Selector
Find embedding models by documented task or category, context, API availability, self-hosting, and input price.
Interactive Tool
Build a Scenario
Ranking priorityChanges weights, not the eligibility gate.
DeploymentUnknown deployment fields fail closed.
| Result | Value | How to read it |
|---|---|---|
| Eligible models | 5 | Models that pass the selected evidence and constraint gates. |
| Ranked shortlist | 5 | Top candidates shown below. |
| Priced candidates | 4 | Models with both input and output price minima. |
| Rank | Model | Fit score | Context | Input / 1M | Workload cost | Published benchmark rank | Providers |
|---|---|---|---|---|---|---|---|
| 1 | Kimi-K2.5Moonshot AI | 48.55 | 262.14K tokens | $0.45 / 1M | $90.00 | Unknown | 2 |
| 2 | Kimi-K2.6Moonshot AI | 46.14 | 262.14K tokens | $0.75 / 1M | $145.00 | Unknown | 2 |
| 3 | Command ACohere | 31.26 | 256K tokens | $2.50 / 1M | $450.00 | Unknown | 1 |
| 4 | Kimi-K3Moonshot AI | 27.5 | 1.05M tokens | $2.85 / 1M | $570.00 | Unknown | 2 |
| 5 | granite-embedding-311m-multilingual-r2IBM | 13.14 | 32.77K tokens | Unknown | Unknown | Unknown | 0 |
Fit scores are transparent screening aids, not universal quality claims. Unknown capability, context, or price fields are never filled by inference.
Inputs
What the Calculation Needs
| Input | How it is used |
|---|---|
| Minimum input length | Largest document or chunk envelope the embedding request needs. |
| Deployment | Require API availability, self-hostability, or either. |
| Maximum input price | Optional per-million-token screening ceiling. |
How to Interpret the Result
Validate retrieval quality, dimensions, multilingual coverage, truncation, and distance metric on your corpus before migrating an index.
Boundaries
What the Result Does Not Prove
- The catalog may not yet expose every vector dimension or benchmark.
- Changing embeddings usually requires a full corpus re-index.
- Input price excludes vector database storage and query serving.
Catalog values retain their source and freshness on the linked model, provider, benchmark, or comparison page. Editable scenario assumptions are not Model Markets measurements.
Continue the Analysis
Related Tools
| Tool | Next question |
|---|---|
| Embedding Cost | Separate one-time corpus embedding cost from recurring refresh and query demand. |
| Model Migration | Calculate payback for switching a production workload instead of comparing token prices alone. |
| Long-Context Models | Filter the catalog by usable request-envelope requirements before comparing cost or benchmark rank. |
Questions
Embedding Model Selector FAQs
What does the Embedding Model Selector calculate?+
Narrow embedding candidates for retrieval, clustering, or semantic search without mixing them with chat models. It returns eligible embedding models with context, deployment options, providers, and known input pricing.
Does the Embedding Model Selector use current model data?+
Where the calculation needs model facts, it uses the current Model Markets catalog snapshot updated 2026-09-02. User-entered assumptions remain clearly editable, and unsupported values stay unknown rather than being inferred.
What should I verify before using the Embedding Model Selector result?+
The catalog may not yet expose every vector dimension or benchmark. Changing embeddings usually requires a full corpus re-index. Input price excludes vector database storage and query serving. Open the linked canonical records and primary sources before making a production or purchasing decision.