Compare model

google / Embedding

GoogleGemini Embedding 2

google/gemini-embedding-2

Latest Gemini embeddings

Best for: Semantic search, RAG and clustering

ShareLinkedInX

Catalog status

Enabled

Research checked

2026-10-06

Benchmark results

3 reported results

Results are source-attributed, not TokenRex measurements. Different tests and thinking variants are not interchangeable.

3 of 3 results

MTEB (Multilingual), Text-Text

Retrieval · Vendor-reported

Test conditions and source

Text-to-text evaluation only (model is natively multimodal but this metric isolates text-text performance). Published source text: Official Google performance table: Gemini Embedding 2 scores 69.9 Mean(Task) on MTEB (Multilingual), Text-Text, vs 68.4 for legacy gemini-embedding-001.

Tested model: google/gemini-embedding-2

Google DeepMind — Gemini Embedding 2 model card · Checked 2026-10-06

69.9mean task score

MTEB (Code)

Retrieval · Vendor-reported

Test conditions and source

MTEB Code benchmark mean task score. Published source text: Official Google performance table: Gemini Embedding 2 scores 84.0 Mean(Task) on MTEB (Code) vs 76.0 for legacy gemini-embedding-001.

Tested model: google/gemini-embedding-2

Google DeepMind — Gemini Embedding 2 model card · Checked 2026-10-06

84mean task score

Image-Text (TextCaps)

Retrieval · Vendor-reported

Test conditions and source

Cross-modal image-to-text retrieval, TextCaps recall@1. Published source text: Official Google performance table: Gemini Embedding 2 Image→Text TextCaps recall@1 = 97.4 (vs 88.1 for legacy multimodalembedding@001, 88.9 Amazon Nova 2, 88.6 Voyage Multimodal 3.5).

Tested model: google/gemini-embedding-2

Google DeepMind — Gemini Embedding 2 model card · Checked 2026-10-06

97.4%

Model specifications

Input and output modalities

Provider-reported formats; API compatibility is detailed below.

Input

Text
Image
Audio
Video
Documents

Output

Embeddings
Exact model ID
google/gemini-embedding-2
Model type
Embedding

Exact token limits are not published in the reviewed sources, so no estimated limits are shown.

Versions and thinking levels

One enabled version is currently listed for this model.

Gemini Embedding 2Selected$0.20 per 1M tokens

Features and tool support

TokenRex’s exposed interface, not every feature advertised by the provider. “Available” describes the implemented interface, not a successful test of every input or tool.

Input through TokenRex

Available

Text or a list of texts on POST /v1/embeddings.

Output through TokenRex

Available

A number vector for each input, ready for search and matching.

Function tools

Not exposed

Embedding endpoint only — it returns vectors and does not run tools.

Sources and verification

Provider specifications describe the upstream model, not a guarantee of every feature through TokenRex. Prices come from the enabled catalog; benchmark results belong to the exact tested model.

Research notes (1)
  • Vendor multimodal embeddings do not guarantee the same input surface through TokenRex.