AI Model Watch
Models / Google

Gemini Embedding 2

GA

Multimodal embedding model (stable/GA). Output is a text embedding vector, not tokens, so there is no output-token price. Input token limit 8,192; output dimension size flexible 128-3072 (recommended 768 / 1536 / 3072). The $0.20/1M figure is the TEXT input rate; the pricing page bills other input types separately - image $0.45, audio $6.50, video $12.00 per 1M (batch is half of each). Release date April 22, 2026 read first-party 2026-08-05 from the Release date column of Google's deprecations table, which carries release dates for CURRENT models too, not only dying ones; Google states 'No shutdown date announced' for it, so retires_on stays null. (Supersedes an earlier note claiming the docs list 'latest update April 2026' rather than a release date - that was true of the model page, which omits the release date; the deprecations table carries it as data.)

Gemini Embedding 2 by Google is tracked live in our LLM catalog, with a 8K (8192-token) context window. It is generally available (GA).

Last verified: 13 Aug 2026 · sourced from official provider documentation

Provider
Google
Status
GA
Input price
$0.2 / 1M tokens
Output price
—
Cached input
—
Blended price
—
Context window
8192 tokens (8K)
Max output
—
Embedding dimensions
3072-dim output vectors
Modality
text, image, video, audio, pdf
Open weights
No — API only
Knowledge cutoff
—
Released
22 Apr 2026
API string
gemini-embedding-2

Source: Google official documentation ↗

See also

Explore more

All Google models→

Prices, context windows and lifecycle status for every Google model we track.

Compare models head-to-head→

Side-by-side price, context and status for the models developers weigh against each other.

FAQ

Gemini Embedding 2 — questions & answers

What is the context window of Gemini Embedding 2?

Gemini Embedding 2 has a 8192-token context window (8K).

Is Gemini Embedding 2 deprecated?

No — Gemini Embedding 2 is generally available (GA) and not currently scheduled for deprecation or retirement in our tracker.

Does Gemini Embedding 2 support image or vision input?

Yes — Gemini Embedding 2 accepts image input. Listed modalities: text, image, video, audio, pdf.

Can I self-host Gemini Embedding 2?

No — Gemini Embedding 2 is proprietary. Its weights are not publicly released, so it is available only through Google's API (or hosted partners), not for self-hosting.

What is the API model string for Gemini Embedding 2?

The API model identifier for Gemini Embedding 2 is "gemini-embedding-2" when calling Google's API.