Google’s Gemini Embedding 2 arrives with native multimodal assist to lower prices and velocity up your enterprise information stack

Yesterday amid a flurry of enterprise AI product updates, Google introduced arguably its most vital one for enterprise prospects: the public preview availability of Gemini Embedding 2, its new embeddings mannequin — a big evolution in how machines characterize and retrieve information throughout totally different media varieties.

Whereas earlier embedding fashions have been largely restricted to textual content, this new mannequin natively integrates textual content, photographs, video, audio, and paperwork right into a single numerical house — lowering latency by as a lot as 70% for some prospects and lowering whole price for enterprises who use AI fashions powered by their very own information to full enterprise duties.

VentureBeat collaborator Sam Witteveen, co-founder of AI and ML coaching firm Crimson Dragon AI, acquired early entry to Gemini Embedding 2 and published a video of his impressions on YouTube. Watch it beneath:

Who wants and makes use of an embedding mannequin?

For many who have encountered the time period “embeddings” in AI discussions however discover it summary, a helpful analogy is that of a common library.

In a conventional library, books are organized by metadata: writer, title, or style. In the “embedding house” of an AI, information is organized by concepts.

Think about a library the place books aren’t organized by the Dewey Decimal System, however by their “vibe” or “essence”. On this library, a biography of Steve Jobs would bodily fly throughout the room to sit subsequent to a technical handbook for a Macintosh. A poem a few sundown would drift towards a images ebook of the Pacific Coast, with all thematically comparable content material organized in lovely hovering “clouds” of books. This is mainly what an embedding mannequin does.

An embedding mannequin takes complicated information—like a sentence, a photograph of a sundown, or a snippet of a podcast—and converts it into an extended listing of numbers referred to as a vector.

These numbers characterize coordinates in a high-dimensional map. If two objects are “semantically” comparable (e.g., a photograph of a golden retriever and the textual content “man’s finest pal”), the mannequin locations their coordinates very shut to one another on this map. As we speak, these fashions are the invisible engine behind:

• Search Engines: Discovering outcomes based mostly on what you imply, not simply the particular phrases you typed.

• Suggestion Methods: Netflix or Spotify suggesting content material as a result of its “coordinates” are close to stuff you already like.

• Enterprise AI: Massive firms use them for Retrieval-Augmented Era (RAG), the place an AI assistant “seems up” an organization’s inside PDFs to reply an worker’s query precisely.

The idea of mapping phrases to vectors dates again to the Fifties with linguists like John Rupert Firth, however the trendy “vector revolution” started in the early 2000s when Yoshua Bengio’s staff first used the time period “phrase embeddings”. The true breakthrough for the {industry} was Word2Vec, launched by a staff at Google led by Tomas Mikolov in 2013. As we speak, the market is led by a handful of main gamers:

• OpenAI: Identified for its widely-used text-embedding-3 sequence.

• Google: With the new Gemini and former Gecko fashions.

• Anthropic and Cohere: Offering specialised fashions for enterprise search and developer workflows.

By transferring past textual content to a natively multimodal structure, Google is trying to create a singular, unified map for the sum of human digital expression—textual content, photographs, video, audio, and paperwork—all residing in the similar mathematical neighborhood.

Why Gemini Embedding 2 is such an enormous deal

Most main fashions are nonetheless “text-first.” In order for you to search a video library, the AI often has to transcribe the video into textual content first, then embed that textual content.

Google’s Gemini Embedding 2 is natively multimodal.

As Logan Kilpatrick of Google DeepMind posted on X, the mannequin permits builders to “deliver textual content, photographs, video, audio, and docs into the similar embedding house”.

It understands audio as sound waves and video as movement instantly, while not having to flip them into textual content first. This reduces “translation” errors and captures nuances that textual content alone may miss.

For builders and enterprises, the “natively multimodal” nature of Gemini Embedding 2 represents a shift towards extra environment friendly AI pipelines.

By mapping all media right into a single 3,072-dimensional house, builders not want separate programs for picture search and textual content search; they will carry out “cross-modal” retrieval—utilizing a textual content question to discover a particular second in a video or a picture that matches a selected sound.

And in contrast to its predecessors, Gemini Embedding 2 can course of requests that blend modalities. A developer can ship a request containing each a picture of a classic automotive and the textual content “What is the engine kind?”. The mannequin does not course of them individually; it treats them as a single, nuanced idea. This permits for a a lot deeper understanding of real-world information the place the “which means” is usually present in the intersection of what we see and what we are saying.

Considered one of the mannequin’s extra technical options is Matryoshka Illustration Studying. Named after Russian nesting dolls, this method permits the mannequin to “nest” the most essential information in the first few numbers of the vector.

An enterprise can select to use the full 3072 dimensions for max precision, or “truncate” them down to 768 or 1536 dimensions to save on database storage prices with minimal loss in accuracy.

Benchmarking the efficiency features of transferring to multimodal

Gemini Embedding 2 establishes a brand new efficiency ceiling for multimodal depth, particularly outperforming earlier {industry} leaders throughout textual content, picture, and video analysis duties.

The mannequin’s most vital lead is present in video and audio retrieval, the place its native structure permits it to bypass the efficiency degradation sometimes related to text-based transcription pipelines.

Particularly, in video-to-text and text-to-video retrieval duties, the mannequin demonstrates a measurable efficiency hole over current {industry} leaders, precisely mapping movement and temporal information right into a unified semantic house.

The technical outcomes present a definite benefit in the following standardized classes:

• Multimodal Retrieval: Gemini Embedding 2 constantly outperforms main textual content and imaginative and prescient fashions in complicated retrieval duties that require understanding the relationship between visible components and textual queries.

• Speech and Audio Depth: The mannequin introduces a brand new commonplace for native audio embeddings, attaining larger accuracy in capturing phonetic and tonal intent in contrast to fashions that rely on intermediate text-transcription.

• Contextual Scaling: In text-based benchmarks, the mannequin maintains excessive precision whereas using its expansive 8,192 token context window, making certain that long-form paperwork are embedded with the similar semantic density as shorter snippets.

• Dimension Flexibility: Testing throughout the Matryoshka Illustration Studying (MRL) layers reveals that even when truncated to 768 dimensions, the mannequin retains a big majority of its 3,072-dimension efficiency, outperforming fixed-dimension fashions of comparable dimension.

What it means for enterprise databases

For the trendy enterprise, information is usually a fragmented mess. A single buyer situation may contain a recorded assist name (audio), a screenshot of an error (picture), a PDF of a contract (doc), and a sequence of emails (textual content).

In earlier years, looking out throughout these codecs required 4 totally different pipelines. With Gemini Embedding 2, an enterprise can create a Unified Information Base. This allows a extra superior type of RAG, whereby an organization’s inside AI does not simply lookup information, however understands the relationship between them no matter format.

Early companions are already reporting drastic effectivity features:

• Sparkonomy, a creator financial system platform, reported that the mannequin’s native multimodality slashed their latency by up to 70%. By eradicating the want for intermediate LLM “inference” (the step the place one mannequin explains a video to one other), they practically doubled their semantic similarity scores for matching creators with manufacturers.

• Everlaw, a authorized tech agency, is utilizing the mannequin to navigate the “high-stakes setting” of litigation discovery. In authorized circumstances the place hundreds of thousands of information should be parsed, Gemini’s capability to index photographs and movies alongside textual content permits authorized professionals to discover “smoking gun” proof that conventional text-search would miss.

Understanding the limits

In its announcement, Google was upfront about a few of the present limitations of Gemini Embedding 2. The brand new mannequin can accommodate vectorization of particular person information that comprise of as many as 8,192 textual content tokens, 6 photographs (in as single batch), 128 seconds of video (2 minutes, 8 seconds lengthy), 80 seconds of native audio (1.34 minutes), and a 6-page PDF.

It is important to make clear that these are enter limits per request, not a cap on what the system can bear in mind or retailer.

Consider it like a scanner. If a scanner has a restrict of “one web page at a time,” it doesn’t suggest you possibly can solely ever scan one web page. it means you may have to feed the pages in one after the other.

• Particular person File Dimension: You can’t “embed” a 100-page PDF in a single name. You need to “chunk” the doc—splitting it into segments of 6 pages or fewer—and ship every section to the mannequin individually.

• Cumulative Information: As soon as these chunks are transformed into vectors, they will all dwell collectively in your database. You possibly can have a database containing ten million 6-page PDFs, and the mannequin will probably be ready to search throughout all of them concurrently.

• Video and Audio: Equally, in case you have a 10-minute video, you’ll break it into 128-second segments to create a searchable “timeline” of embeddings.

Licensing, pricing, and availability

As of March 10, 2026, Gemini Embedding 2 is formally in Public Preview.

For builders and enterprise leaders, this implies the mannequin is accessible for quick testing and manufacturing integration, although it is nonetheless topic to the iterative refinements typical of “preview” software program before it reaches Basic Availability (GA).

The mannequin is deployed throughout Google’s two major AI gateways, every catering to a unique scale of operation:

• Gemini API: Focused at fast prototyping and particular person builders, this path provides a simplified pricing construction.

• Vertex AI (Google Cloud): The enterprise-grade setting designed for enormous scale, providing superior safety controls and integration with the broader Google Cloud ecosystem.

It is also already built-in with the heavy hitters of AI infrastructure: LangChain, LlamaIndex, Haystack, Weaviate, Qdrant, and ChromaDB.

In the Gemini API, Google has launched a tiered pricing mannequin that distinguishes between “commonplace” information (textual content, photographs, and video) and “native” audio.

• The Free Tier: Builders can experiment with the mannequin for free of charge, although this tier comes with charge limits (sometimes 60 requests per minute) and makes use of information to enhance Google’s merchandise.

• The Paid Tier: For production-level quantity, the price is calculated per million tokens. For textual content, picture, and video inputs, the charge is $0.25 per 1 million tokens.

• The “Audio Premium”: As a result of the mannequin natively ingests audio information with out intermediate transcription—a extra computationally intensive activity—the charge for audio inputs is doubled to $0.50 per 1 million tokens.

For giant-scale deployments on Vertex AI, the pricing follows an enterprise-centric “Pay-as-you-go” (PayGo) mannequin. This permits organizations to pay for precisely what they use throughout totally different processing modes:

• Flex PayGo: Greatest for unpredictable, bursty workloads.

• Provisioned Throughput: Designed for enterprises that require assured capability and constant latency for high-traffic purposes.

• Batch Prediction: Excellent for re-indexing huge historic archives, the place time-sensitivity is decrease however quantity is extraordinarily excessive.

By making the mannequin out there by these numerous channels and integrating it natively with libraries like LangChain, LlamaIndex, and Weaviate, Google has ensured that the “switching price” for companies is not only a matter of value, however of operational ease. Whether or not a startup is constructing its first RAG-based assistant or a multinational is unifying a long time of disparate media archives, the infrastructure is now dwell and globally accessible.

As well as, the official Gemini API and Vertex AI Colab notebooks, which comprise the Python code mandatory to implement these options, are licensed below the Apache License, Model 2.0.

The Apache 2.0 license is extremely regarded in the tech neighborhood as a result of it is “permissive.” It permits builders to take Google’s implementation code, modify it, and use it in their very own business merchandise with out having to pay royalties or “open supply” their very own proprietary code in return.

How enterprises ought to reply: migrate to Gemini 2 Embedding or not?

For Chief Knowledge Officers and technical leads, the resolution to migrate to Gemini Embedding 2 hinges on the transition from a “text-plus” technique to a “natively multimodal” one.

In case your group at present depends on fragmented pipelines — the place photographs and movies are first transcribed or tagged by separate fashions before being listed — the improve is possible a strategic necessity.

This mannequin eliminates the “translation tax” of utilizing intermediate LLMs to describe visible or auditory information, a transfer that companions like Sparkonomy discovered decreased latency by up to 70% whereas doubling semantic similarity scores. For companies managing huge, numerous datasets, this is not only a efficiency enhance; it is a structural simplification that reduces the variety of factors the place “which means” will be misplaced or distorted.

The hassle to change from a text-only basis is decrease than one may count on due to what early customers describe as wonderful “API continuity”.

As a result of the mannequin integrates with industry-standard frameworks like LangChain, LlamaIndex, and Vector Search, it might probably usually be “dropped into” current workflows with minimal code adjustments. Nevertheless, the actual price and power funding lies in re-indexing. Transferring to this mannequin requires re-embedding your current corpus to guarantee all information factors exist in the similar 3,072-dimensional house.

Whereas this is a one-time computational hurdle, it is the prerequisite for unlocking cross-modal search—the place a easy textual content question can instantly “see” into your video archives or “hear” particular buyer sentiment in name recordings.

The first trade-off for information leaders to weigh is the stability between high-fidelity retrieval and long-term storage economics. Gemini Embedding 2 addresses this instantly by Matryoshka Illustration Studying (MRL), which permits you to truncate vectors from 3072 dimensions down to 768 with out a linear drop in high quality.

This provides CDOs a tactical lever: you possibly can select most precision for high-stakes authorized or medical discovery—as seen in Everlaw’s 20% elevate in recall—whereas using smaller, extra environment friendly vectors for lower-priority advice engines to preserve cloud storage prices in verify.

In the end, the ROI is present in the “elevate” of accuracy; in a panorama the place an AI’s worth is outlined by its context, the capability to natively index a 6-page PDF or 128 seconds of video instantly right into a information base offers a depth of perception that text-only fashions merely can not replicate.