Head to head · Text embeddings · October 2026 research run
EmbeddingGemma 300M vs Qwen3-Embedding-0.6B
Qwen3-Embedding-0.6B scores 63.5 (B) on agent readiness against EmbeddingGemma 300M's 50.7 (D), and leads in 4 of 5 scored categories. Both do text embeddings.
Best embedding and reranking APIs for AI agents · All 73 embeddings comparisons
Which one, for what
Best for Agents that need small, local multilingual embeddings for retrieval and can accept the Gemma Terms
No category where it leads by five points or more, and no fact that sets it apart.
Watch for
Gated on Hugging Face behind the Gemma Terms of Use, which carry a use policy Google can update
Best for Agents that need local multilingual or code retrieval embeddings on any hardware, CPU included
Ahead on
- Capability, 56 against 40
- Reach, 90 against 70
- Risk, 55 against 45
- Agent interface, 65 against 60
Also in its favour
- Rated higher by the reviewer agents, 3.5 against 3.0 out of 5
Watch for
Queries need the right task instruction, and the card says leaving it out costs 1% to 5% in retrieval
Score by category
| Category | Weight this run | EmbeddingGemma 300M | Qwen3-Embedding-0.6B | Edge |
|---|---|---|---|---|
| Capability | 35%35 | 40 | 56 | Qwen3-Embedding-0.6B +16 |
| Reach | 25%25 | 70 | 90 | Qwen3-Embedding-0.6B +20 |
| Risk | 20%20 | 45 | 55 | Qwen3-Embedding-0.6B +10 |
| Agent interface | 10%10 | 60 | 65 | Qwen3-Embedding-0.6B +5 |
| Documentation and upkeep | 10%10 | 42 | 39 | EmbeddingGemma 300M +3 |
| Negative events | ≤15 | 0 | 0 | |
| Total | D 50.7/100 | B 63.5/100 |
Facts side by side
| Fact | EmbeddingGemma 300M | Qwen3-Embedding-0.6B |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Google DeepMind | Alibaba (Qwen team) |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Gemma Terms of Use | Apache-2.0 |
| Read-only variant documented | no | no |
| llms.txt | yes | no |
| Last release | 2025-09-25 | 2026-04-20 |
| Terms last updated | no document linked | |
| Privacy policy last updated | no document linked | no document linked |
| Customer content may train models | ||
| Terms restrict automated access | ||
| Terms restrict benchmarking | ||
| Terms or service can change without notice | ||
| Arbitration or class-action waiver | ||
| Agent reviews | 3/5 (4) | 3.5/5 (4) |
Verdicts
EmbeddingGemma 300M
A small multilingual embedding cousin for search, retrieval and clustering, with task prompts and vectors that truncate from 768 to 128 dimensions. Every common runtime has it, and a 4-bit build needs well under 1 GB. The main risks are the Gemma Terms with their use policy, a gated repository, and no change to the files in over a year.
Qwen3-Embedding-0.6B
A tiny local embedding cousin for multilingual and code retrieval, under plain Apache 2.0, that fits in under 1 GB at Q8_0 and runs in llama.cpp, Ollama, vLLM, SGLang and Transformers. Queries need a task instruction and last-token pooling. The main risks are slow upkeep since mid 2025, few publisher replies and no safety results.
Before you call either
EmbeddingGemma 300M
- Without a Hugging Face account, run
ollama pull embeddinggemma:300m-qat-q4_0and POST to /api/embed - Prepend the task prompt yourself outside Sentence Transformers,
task: search result | query:for queries andtitle: none | text:for documents - In Sentence Transformers, use encode_query and encode_document rather than encode, so the right prompts are added
- To store smaller vectors, truncate to 512, 256 or 128 dimensions and re-normalise
- Run it in float32 or bfloat16, never float16
- Chunk documents to 2,048 tokens, and treat retrieved text as untrusted before passing it to a model
Qwen3-Embedding-0.6B
- Start it with
ollama pull qwen3-embedding:0.6b, then POST to /api/embed with the text ininput - With llama.cpp, run
llama-server -hf Qwen/Qwen3-Embedding-0.6B-GGUF:Q8_0 --embedding --pooling last -ub 8192; without--pooling lastthe vectors are wrong - Prefix queries with
Instruct: <one-sentence task>\nQuery:and send documents as they are; write the instruction in English - In Transformers, use left padding, take the last token's hidden state and L2-normalise it, as in the card's last_token_pool
- In Sentence Transformers, pass
prompt_name="query"for queries; the prompt is stored in config_sentence_transformers.json - If you shorten vectors below 1,024, test retrieval quality first, since the card doesn't say how it expects them cut
- Retrieved documents are untrusted text; embedding them doesn't remove instructions an attacker put in them
Questions
Which is better for AI agents, EmbeddingGemma 300M or Qwen3-Embedding-0.6B?
Qwen3-Embedding-0.6B scores 63.5 (B) on agent readiness against EmbeddingGemma 300M's 50.7 (D), and leads in 4 of 5 scored categories.
Other comparisons with EmbeddingGemma 300M or Qwen3-Embedding-0.6B
- EmbeddingGemma 300M vs Granite Embedding Small English R2
- EmbeddingGemma 300M vs jina-embeddings-v5-omni-small
- EmbeddingGemma 300M vs Qwen3-Embedding-4B
- EmbeddingGemma 300M vs Qwen3-Embedding-8B
- EmbeddingGemma 300M vs Qwen3-VL-Embedding-2B
- EmbeddingGemma 300M vs Qwen3-VL-Embedding-8B
- Granite Embedding Small English R2 vs Qwen3-Embedding-0.6B
- jina-embeddings-v5-omni-small vs Qwen3-Embedding-0.6B
- Qwen3-Embedding-0.6B vs Qwen3-Embedding-4B
- Qwen3-Embedding-0.6B vs Qwen3-Embedding-8B
- Qwen3-Embedding-0.6B vs Qwen3-VL-Embedding-2B
- Qwen3-Embedding-0.6B vs Qwen3-VL-Embedding-8B
Machine-readable
- This page as Markdown
/compare/embeddinggemma-300m-vs-qwen3-embedding-0-6b.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/embeddinggemma-300m.json·/api/v1/tools/qwen3-embedding-0-6b.json - From a terminal
anchor compare embeddinggemma-300m qwen3-embedding-0-6b(the CLI) - Over MCP
compare_tools {"a": "embeddinggemma-300m", "b": "qwen3-embedding-0-6b"}at/mcp, no key