EmbeddingGemma
The most recommended of 8 embedding models
Open embedding model people praise for quality, low RAM use and its Apache 2.0 license. Critics note weaker accuracy than cloud models for speech and some tasks.
Open embedding model people praise for quality, low RAM use and its Apache 2.0 license. Critics note weaker accuracy than cloud models for speech and some tasks.
31 people gave an opinion
- recommend it
- 27
- criticize it
- 4
- approval share
- 87%
Recommended in 3 of 13 months
- November 2025: recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: not recommended
- May 2026: recommended
- June 2026: not recommended
- July 2026: not recommended
- August 2026: not recommended
- September 2026: not recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 27 people recommend itNeeds 10 or more, met.
- 6.8 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 3 of 13 monthsNeeds 4 or more, not met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Strong benchmarks and quality
“Better quality, less RAM usage, faster embeddings, I'm fucking pumped for it.”
Comment from 7 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” “has incredible benchmarks and a 2k context window”
Comment from 28 Nov 2025 in “28M Hacker News comments as vector embedding search dataset” “which has incredible benchmarks”
Comment from 29 Nov 2025 in “28M Hacker News comments as vector embedding search dataset” - Comment from 13 Oct 2026 in “We turned EmbeddingGemma 2 into a speech-to-text model”
Small and efficient for its quality
“270M for text only is great compared to older embedding models”
Comment from 6 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” “embeddiggemma because it's crazy efficient”
Comment from 5 May 2026 in “Ask HN: Best Embedding Models?” “Better quality, less RAM usage, faster embeddings, I'm fucking pumped for it.”
Comment from 7 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” Open weights under Apache 2.0
“Hats off to google for offering OSS (or at least open weights + license) a model”
Comment from 6 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” “I really appreciate that EmbeddingGemma 2 is under the Apache 2.0 license.”
Comment from 6 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” Multimodal text, image and video use
“And if your indexing software uses it correctly you can also do video deposition searches”
Comment from 7 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model” “It's also very neat that this can be used for "Jev"-like tasks with text and image.”
Comment from 6 Oct 2026 in “EmbeddingGemma 2: An open, lightweight multimodal embedding model”
What people warn about
Each mark is one comment
Transcription accuracy below cloud models
“WER is not good enough for professional use, fine for hobby projects”
Comment from 13 Oct 2026 in “We turned EmbeddingGemma 2 into a speech-to-text model” “The accuracy is good, but not in the range of cloud models”
Comment from 13 Oct 2026 in “We turned EmbeddingGemma 2 into a speech-to-text model” “the overall transcription quality is very low”
Comment from 13 Oct 2026 in “We turned EmbeddingGemma 2 into a speech-to-text model”
Other embedding models people recommend
The four most recommended other embedding models, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| EmbeddingGemmaThis page | 27 | 4 | 87% | 3 | — | — |
| Ternlight | 17 | 3 | 85% | 1 | — | — |
| CLIP by OpenAI | 10 | 5 | 67% | 6 | — | Approved |
| Gemini Embedding 2 | 9 | 3 | 75% | 3 | — | — |
| Qwen3-Embedding | 8 | 1 | 89% | 4 | — | — |
What people said
32 of the 33 comments that mention it, newest first. Each links to the original comment.
“It's been absolutely blowing my mind.”
“seems far more capable than foundation models”
“rather than ones that "support" it like embeddinggemma”
“nice work, im experimenting it with Urdu”
“great performance achieved”
“It worked well on domains I trained it for... it never generalized. Out of the trained domains, it was just above chance.”
“the overall transcription quality is very low”
“People aren't paying enough attention to how great EmbeddingGemma 2 and Gemma 4 are.”
“The accuracy is good, but not in the range of cloud models”
“this seems to target English only”
“it took 6.75 seconds to transcribe it using this software”
“WER is not good enough for professional use, fine for hobby projects”
“Kudos for it and I will try it out!”
“Interesting as a proof of concept.”
“Per https://ai.google.dev/gemma/docs/embeddinggemma/model_card_2... the audio data mentions speech, environment sounds”
“And if your indexing software uses it correctly you can also do video deposition searches”
“not better or faster if one only needs text embeddings for retrieval or clustering”
“Better quality, less RAM usage, faster embeddings, I'm fucking pumped for it.”
“failed. It said the request does not involve a payment, charge, or refund”
“I suspect that this compares favorably in 9X% of use cases”
“It's also very neat that this can be used for "Jev"-like tasks with text and image.”
“you don’t get to shrink the model weights alongside the lower dimensional embeddings, unfortunately”
“this seems like a really nice embedding model”
“Hats off to google for offering OSS (or at least open weights + license) a model”
“I really appreciate that EmbeddingGemma 2 is under the Apache 2.0 license.”
“270M for text only is great compared to older embedding models”
“embeddiggemma because it's crazy efficient”
“which has incredible benchmarks”
“Google could at any point in the future decide that the thing you are building is now a restricted use”
“I got way better results out of the nomic model.”
“has incredible benchmarks and a 2k context window”
“But when using custom task types, the vector arithmetic worked as expected.”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.