Elastic Brings Jina Embeddings v5 to Microsoft Foundry

Elastic's Jina v5 text and omni embedding models now deploy as dedicated endpoints inside customers' Microsoft Azure environments.

Updated on Nov. 20, 2026 in Artificial Intelligence

Elastic Brings Jina Embeddings v5 to Microsoft Foundry

Elastic has made its jina-embeddings-v5-text and jina-embeddings-v5-omni models available in the Microsoft Foundry model catalog. The models deploy as dedicated endpoints in each customer's own Microsoft Azure environment.

Why it matters

The release gives developers access to highly accurate semantic search and retrieval without the compute costs of larger models. Running the models inside a customer's own Azure account keeps deployment within an existing cloud setup.

The Jina v5 family includes two text models and two omni models, each offered in two parameter sizes, with the text models supporting more than 100 languages. The omni models add image, audio, and video support and are reported to outperform models up to 20 times larger on image benchmarks, though the source does not provide independent verification of these results.

The players

Elastic

Elastic is the company behind Elasticsearch, a search and analytics platform used by a large share of Fortune 100 companies.

The details

The jina-embeddings-v5-text and jina-embeddings-v5-omni models are now listed in the Microsoft Foundry model catalog. Developers run them as dedicated endpoints in their own Azure accounts. The omni models allow a single query to search across images, audio, and video within one vector space.

Timeline

  1. November 19, 2026: Elastic announced the models' availability in Microsoft Foundry.

The Tech Race

The listing places Elastic's Jina embedding models alongside other models offered through the Microsoft Foundry catalog, extending the choice available to Azure customers building retrieval systems. It reflects cloud providers competing to host a wider range of third-party AI models on their platforms.

Developers building search or retrieval features can deploy these embedding models in their own Azure environments rather than relying on external hosted endpoints. The omni models let a single query search across images, audio, and video, which may simplify building multimodal search tools.

The takeaway

Smaller embedding models can deliver strong retrieval accuracy at lower compute cost, so teams weighing model size against infrastructure spend have a new option to evaluate. Testing candidate models against your own data remains the practical way to judge fit.

Further reading

For broader context on how such models are changing search and retrieval, see Artificial Intelligence coverage on Wisevoter.

Source note: This article includes information reported by HPCwire.