Model Tracker
EmbeddingGemma 2
EmbeddingGemma 2 is an AI model made by Google. Google released it on 6 October 2026. It is generally available. Its context window is 8,192 tokens.
En anglais
EmbeddingGemma 2 is an AI model made by Google. Google released it on 6 October 2026. It is generally available. Its context window is 8,192 tokens. Its weights are open. Its licence is Apache 2.0. Its knowledge cutoff is January 2025.
- Released
- 6 October 2026 — gemma.ai, read 18 November 2026
- Status
- Generally available — blog.google, read 18 November 2026
- Takes
- Not confirmed: two readings disagree: text, image, audio, video, code and text, image, audio, video
- Returns
- Not confirmed: The model returns embeddings (vectors), but the register's allowed output values have no embedding or vector category; assigning one of the listed modalities would be unsupported.
- Open weights
- Yes — blog.google, read 18 November 2026
- Licence
- Apache 2.0 — gemma.ai, read 18 November 2026
- Context window
- 8,192 tokens — gemma.ai, read 18 November 2026
- Longest answer
- Not confirmed: The maker describes an embedding model that returns vectors and does not publish a longest-answer token limit.
- Knowledge cutoff
- January 2025 — gemma.ai, read 18 November 2026
- API price, input
- Not confirmed: The maker's pages do not publish a standard API input price per 1 million tokens for this model.
- API price, output
- Not confirmed: The maker's pages do not publish a standard API output price per 1 million tokens for this model.
- Where
- Not confirmed: its line was not found on its page, and a second reading did not confirm it
- Parameters
- 740,000,000 — gemma.ai, read 18 November 2026
- Model card
- https://gemma.ai/models/embeddinggemma/model-card-v2/ — gemma.ai, read 18 November 2026
In the news
- Google details EmbeddingGemma 2’s multimodal design in report (17 November 2026)
- Google Announces Offline Foresight Meeting Companion for Mac (9 October 2026)
- Google Releases 740M-Parameter EmbeddingGemma 2 for Edge AI (6 October 2026)
Questions
When was EmbeddingGemma 2 released?
Google released EmbeddingGemma 2 on 6 October 2026.
Is EmbeddingGemma 2 still available?
Yes: it is generally available, as Google states it.
What is the context window of EmbeddingGemma 2?
The context window of EmbeddingGemma 2 is 8,192 tokens, as Google prints it.
Are the weights of EmbeddingGemma 2 open?
Yes: Google publishes the weights of EmbeddingGemma 2 for download.
What is the licence of EmbeddingGemma 2?
The licence of EmbeddingGemma 2 is Apache 2.0, as Google names it.
What is the knowledge cutoff of EmbeddingGemma 2?
The knowledge cutoff of EmbeddingGemma 2 is January 2025.
How many parameters does EmbeddingGemma 2 have?
EmbeddingGemma 2 has 740,000,000 parameters, as Google publishes it.
More from Google
- CodeGemma
- DiffusionGemma
- Gemini 1.0 Pro Vision
- Gemini 1.5 Flash
- Gemini 1.5 Flash-8B
- Gemini 1.5 Pro
- Gemini 2.0 Flash
- Gemini 2.0 Flash Experimental
- Gemini 2.0 Flash-Lite
- Gemini 2.5 Computer Use
- Gemini 2.5 Deep Think
- Gemini 2.5 Flash
- Gemini 2.5 Flash Image
- Gemini 2.5 Flash Live
- Gemini 2.5 Flash TTS
- Gemini 2.5 Flash-Lite
- Gemini 2.5 Pro
- Gemini 2.5 Pro TTS
- Gemini 3 Flash
- Gemini 3 Pro
- Gemini 3.1 Flash Live
- Gemini 3.1 Flash TTS
- Gemini 3.1 Flash-Lite
- Gemini 3.1 Pro
Sources
- EmbeddingGemma 2 model card (read 18 November 2026)
- Releases (read 18 November 2026)
- EmbeddingGemma 2: an open, lightweight multimodal embedding model (read 18 November 2026)