Data as of November 21, 2026 · How the score is built
EmbeddingGemma 2
Decision readingEmbeddingGemma 2 is tracked, but not publicly ranked yet. The profile exposes 0 sourced benchmark rows and leaves unsupported fields blank until a published record exists.
Email me when its price, rank or availability changes.
Benchmark ledger
Every result says who ran the test. For this model, Google ran all 8.
| Benchmark | Score | On a 0–100 scale | Who ran it | Use in ranking |
|---|---|---|---|---|
| MTEB (code, v1)Mean(Task), NDCG@10 · Full precision, 768 dimensions | 78.68 | Reported by Google | Shown only | |
| MSEB (Retrieval)Mean(Task), MRR@10 · Full precision, 768 dimensions | 69.54 | Reported by Google | Shown only | |
| MMEB v2, VisDocMean(Task), NDCG@5 · Full precision, 768 dimensions | 67.84 | Reported by Google | Shown only | |
| MIEB (lite)Mean(TaskType), Multiple · Full precision, 768 dimensions | 65.67 | Reported by Google | Shown only | |
| MTEB (multilingual, v2)Mean(Task), multiple metrics · Full precision, 768 dimensions | 61.36 | Reported by Google | Shown only | |
| MMEB v2, ImageMean(Task), Hit@1 · Full precision, 768 dimensions | 57.28 | Reported by Google | Shown only | |
| MMEB v2, VideoMean(Task), Hit@1 · Full precision, 768 dimensions | 50.67 | Reported by Google | Shown only | |
| MAEBMean(Task), Multiple · Full precision, 768 dimensions | 49.87 | Reported by Google | Shown only |
Outlined bars are lab-reported. Filled bars are benchmark-run or independent.
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Embedding
Spec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- Weights model ID
- google/embeddinggemma-2Google EmbeddingGemma 2 model card
- Context window
- 8192 tokensGoogle EmbeddingGemma 2 model card
- Maximum output
- Not sourced yet
- Knowledge cutoff
- Not sourced yet
- Input modalities
- text, image, video, audioGoogle EmbeddingGemma 2 model card
- Output modalities
- embeddingGoogle EmbeddingGemma 2 model card
- Parameters
- Not sourced yet
- Availability
- Hugging Face · Kaggle · Self-hosted inferenceGoogle EmbeddingGemma 2 model card
- Cloud regions
- Not tracked yet
- Lifecycle
- activeGoogle EmbeddingGemma 2 model card
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Not documented in the pricing recordGoogle EmbeddingGemma 2 model card
- Self-host
- Open weights available; hardware estimate not sourced
- Rate limits
- Not tracked yet
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
We track EmbeddingGemma 2, but no weighted text-model benchmark result is published on the site yet. This page shows the metadata and separate protocol evidence we can verify now; a BenchLM score will appear only if compatible public evaluations land.
EmbeddingGemma 2 is a open weight model with a 8192 tokens context window. No explicit reasoning mode is documented in this profile.
Google released EmbeddingGemma 2 on October 6, 2026. It maps text, code, images, video, and audio into 768-dimensional vectors, with 512, 256, and 128 dimensions also supported. The model has 740M parameters: a 270M text model, a 170M vision encoder, and a 300M audio encoder. Google reports an 8,192-token context window and 100+ supported languages. At full precision and 768 dimensions, Google's results on MTEB multilingual v2, MTEB code v1, MIEB lite, MMEB v2 Image, VisDoc, and Video, MSEB Retrieval, and MAEB are listed in the benchmark ledger. These embedding evaluations are outside the current benchmark schema and do not establish general LLM scores.
The profile has no source-displayable benchmark row yet.
Last updated November 21, 2026. Runtime fields remain blank until a sourced snapshot exists.
Deployment options
Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.
Published weights are available, but BenchLM does not yet have a sourced parameter and VRAM profile for this exact model. Hardware cost estimates stay unavailable until that sizing record is complete.
Estimate VRAM from known parametersQuestions
How does EmbeddingGemma 2 perform overall in AI benchmarks?
EmbeddingGemma 2 does not have any source-displayable benchmark rows yet, so this profile does not assign a public score or rank. Documented specifications remain visible, while score-led charts and claims stay unavailable until a published evaluation can be attached to the exact model.
Is EmbeddingGemma 2 open source?
EmbeddingGemma 2 is an open-weight model from Google. Its weights can be downloaded for local or hosted deployment, subject to the published license. Open weight does not automatically mean open source: training data and training code may remain private, and commercial restrictions can still apply.
Does EmbeddingGemma 2 have full benchmark coverage on BenchLM?
No. EmbeddingGemma 2 currently has 0 source-displayable rows. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.
What is the context window size of EmbeddingGemma 2?
EmbeddingGemma 2 has a documented context window of 8192 tokens. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.