Elastic announced the availability of jina-embeddings-v5-text, a family of two small, Elasticsearch-native multilingual embedding models at 0.2B and 0.6B parameters that deliver state-of-the-art performance across key search and semantic tasks.
Despite their compact size, they outperform significantly larger models with 7B to 14B parameters and achieve best-in-class results on the MMTEB (Multilingual MTEB) benchmark among models of comparable size and purpose. Their small footprint enables outstanding hybrid search at lower infrastructure cost, faster query response, and new deployment scenarios where memory and compute budgets are tight - including edge devices and resource-constrained environments.
jina-embeddings-v5-text are available through multiple channels: as open-weight models on HuggingFace for self-hosted deployment via vLLM, llama.cpp, or MLX, and on Elastic Inference Service (EIS), a GPU-accelerated inference-as-a-service that makes it easy to run fast, high-quality inference without complex setup. By bringing the Jina v5 family to EIS, users get a complete data platform that consolidates state-of-the-art multilingual embedding models, a high-performance vector database, and more into one unified enterprise stack across cloud and on-premises.
“Vector search, RAG, and AI agents depend on high-quality retrieval,” said Steve Kearns, general manager, Search, Elastic. “With the addition of the Jina v5’s multilingual embeddings, Elasticsearch continues to be the platform of choice for end-to-end context engineering.”
The family includes two models, jina-embeddings-v5-text-small (239M parameters) and jina-embeddings-v5-text-nano (677M parameters). Both models are optimized for four common tasks in search and agentic applications:
- Retrieval: Allowing users to query with natural language and find the most relevant documents
- Text Matching: Allowing users to find duplicates in their data, and align paraphrases or translations
- Classification: Allowing users to categorize documents, detect sentiments, and find anomalies
- Clustering: Allowing users to group documents by topic, subject, or meaning
The Jina v5 models are now available through the Elastic Inference Service (EIS) on Elastic Cloud Serverless and Elastic Cloud Hosted. All Elastic Cloud Trials include access to EIS.
These models are also available via an online API, and available for local hosting via vLLM, llama.cpp and MLX. Detailed instructions can be found on Hugging Face.
The Latest
IT organizations have historically measured success by how quickly they can respond when something goes wrong. The entire discipline of Incident Management has been optimized around mean time to resolution, first-response SLAs and ticket closure rates. But new research suggests that even though this is a well-executed playbook, it's no longer enough to retain customers ...
Production incidents rarely announce themselves as database problems. They appear as slow transactions, timeouts, rising response times, or an application struggling under a workload it previously handled. APM provides an essential starting point. It can identify a slow transaction path, highlight an affected service, and show that a database dependency is consuming more time than expected. But identifying the database as part of the problem is not the same as explaining what is happening inside it ...
Cloud teams are under constant pressure to reduce spend without slowing development or increasing operational risk. They are deploying autoscalers, rightsizing workloads, enforcing resource requests, reviewing utilization dashboards, and building FinOps processes around cloud-native environments. Yet the results often disappoint ...
Ask most IT leaders about their biggest concern with AI and you'll hear the same answer: hallucinations ... Today, however, the conversation has shifted ... As organizations move beyond chatbots and experiments, they are increasingly deploying AI agents that perform multi-step tasks. These systems retrieve documents, query databases, call APIs, generate reports, write code, and make recommendations. The issue is not whether the model can reason. The issue is whether the organization can see, verify, and govern the decisions being made along the way ...
While organizations want to take control of their telemetry, building telemetry pipelines from scratch can be a very daunting, complicated task, even when leveraging open-source standards like OpenTelemetry. It requires specialized knowledge across distributed systems, data engineering, and security. This fragmented approach across systems causes higher operational costs; it puts a strain on resources and reduces efficiency as teams have to work with different interfaces and processes ...
For decades, enterprise networks were designed around a simple assumption: work happened inside the office. Applications lived in centralized data centers, employees connected through internal infrastructure, and security focused on protecting the perimeter that surrounded everything ... But the way organizations operate today bears little resemblance to that environment. Cloud platforms host critical applications, employees connect from homes and airports as often as they do from offices, and partners collaborate through shared systems that exist far beyond corporate walls. In short, the corporate network no longer resembles the environment it was designed to protect ...
As an analyst who researches how IT organizations design, build, and operate their networks, I find that network data is a constant source of pain. Network teams struggle with data quality, fragmentation, authority, access, and trust. And these issues undermine everything they try to do. Here are the numbers: Only 45% of network teams are completely confident in the accuracy of their network source of truth, which documents the intent of their network ...
The 2026 Global Data Center Survey from Uptime Institute reveals an industry navigating workforce constraints, escalating outage expenses, even as rising costs remain the top concern for management teams ...
The next observability gap may not be in the code. It may be under the rack. That sounds strange until you think about how AI incidents actually feel in the middle of an investigation ... The application dashboard may be accurate. It may also be stopping at the wrong boundary. AI systems depend on software, but they also depend on a dense physical stack: racks, power paths, thermal margin, maintenance activity and, in many environments, liquid cooling. Those physical dependencies can change slowly before they look like a software incident ...
Certificate expiration is the rare outage you can see coming. Every TLS certificate carries the date it stops working, so the moment it will begin breaking connections is knowable in advance. That's what makes an expired certificate such a frustrating way to lose a service. What's changing now is how often that date comes around ...