Cohere / Embed
Compact Multilingual embeddings for retrieval, similarity and classification, with text and image support in the current API.
Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.
Per-input text-token limit; batching does not multiply the limit for an individual text.
The current catalog explicitly lists Embed Jobs for this exact model. Asynchronous job limits and prices require a separate review.
Original text-embedding family launch; later image support is not backdated to this date.
Fixed vector dimension for this release.
Detailed Embed reference lists all three; the shorter catalog table lists cosine only for Embed 3.
float is the default. Binary representations pack eight dimensions into each output integer; base64 encodes concatenated float bytes.
Image data URIs. Mixed text/image inputs are documented for Embed 4 and later; do not infer a direct PDF upload endpoint from PDF-page examples.
Request batching is distinct from asynchronous Embed Jobs. Token limits still apply per input.
These are embedding API controls, not text-generation sampling parameters.
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
embed-multilingual-light-v3.0Endpoint pricing has not been collected. Check the provider for applicable rates.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Not collectedDeployment-specific identifier; do not substitute the Cohere API model selector.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
cohere.embed-multilingual-light-v3.0On-demand retirement. Dedicated table independently lists the same date; active dedicated deployment exceptions remain.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
cohere.embed-multilingual-light-image-v3.0Endpoint pricing has not been collected. Check the provider for applicable rates.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
The system processes eligible input and retains reusable state.
Reuse depends on matching input, session state and the cache policy.
Reusing state can reduce repeated work. Any charges follow the billing rules below.
embed-multilingual-light-v3.0No model-specific embedding cache contract verified. A shared API cached_tokens schema field is insufficient evidence of support.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026No model-specific embedding cache contract verified. A shared API cached_tokens schema field is insufficient evidence of support.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026cohere.embed-multilingual-light-v3.0No model-specific embedding cache contract verified. A shared API cached_tokens schema field is insufficient evidence of support.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026cohere.embed-multilingual-light-image-v3.0No model-specific embedding cache contract verified. A shared API cached_tokens schema field is insufficient evidence of support.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The provider’s exact-model repository publishes tokenizer/configuration files, not model weights. Its model card calls it a tokenizer repository and directs inference to the Cohere API, SageMaker or sales-arranged private deployment. Public tokenizer files do not make the model open-weight.
Classification evidence (opens in a new tab)