Head to head · Text embeddings · October 2026 research run
Qwen3-Embedding-0.6B vs Qwen3-Embedding-4B
Qwen3-Embedding-4B scores 72.9 (BB) on agent readiness against Qwen3-Embedding-0.6B's 63.5 (B), and leads in 2 of 5 scored categories. Qwen3-Embedding-0.6B leads on documentation and upkeep. 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 local multilingual or code retrieval embeddings on any hardware, CPU included
Ahead on
- Documentation and upkeep, 39 against 21
Watch for
Queries need the right task instruction, and the card says leaving it out costs 1% to 5% in retrieval
Best for Agents that need local multilingual text or code retrieval on a laptop, a small GPU or a CPU, with a plain licence.
Ahead on
- Capability, 81 against 56
- Reach, 100 against 90
Also in its favour
- Agent-ready, a grade of BB or better
Watch for
No change to the files or reference code since August 2025, and no reply from Qwen to any of its 16 questions
Score by category
| Category | Weight this run | Qwen3-Embedding-0.6B | Qwen3-Embedding-4B | Edge |
|---|---|---|---|---|
| Capability | 35%35 | 56 | 81 | Qwen3-Embedding-4B +25 |
| Reach | 25%25 | 90 | 100 | Qwen3-Embedding-4B +10 |
| Risk | 20%20 | 55 | 55 | even |
| Agent interface | 10%10 | 65 | 65 | even |
| Documentation and upkeep | 10%10 | 39 | 21 | Qwen3-Embedding-0.6B +18 |
| Negative events | ≤15 | 0 | 0 | |
| Total | B 63.5/100 | BB 72.9/100 |
Facts side by side
| Fact | Qwen3-Embedding-0.6B | Qwen3-Embedding-4B |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Alibaba (Qwen team) | Alibaba (Qwen team) |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Apache-2.0 | Apache-2.0 |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-04-20 | 2025-08-03 |
| Terms last updated | no document linked | 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 | 3.5/5 (4) | 3.5/5 (4) |
Verdicts
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.
Qwen3-Embedding-4B
A small multilingual text embedding cousin for retrieval and code search that runs in about 3 GB from Alibaba's own 4-bit GGUF, loads in every major runtime under Apache 2.0 and has two hosts, DeepInfra and SiliconFlow. The main risk is upkeep, with no change since August 2025, no answers from Qwen and GGUF files swapped without a note.
Before you call either
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
Qwen3-Embedding-4B
- Quickest start is
ollama pull qwen3-embedding:4b, about 2.5 GB, then POST to /api/embed - With llama.cpp use
llama-server -hf Qwen/Qwen3-Embedding-4B-GGUF:Q4_K_M --embedding --pooling last -ub 8192, the flags the GGUF card gives - Prefix each query with
Instruct: <one-sentence task>\nQuery:and embed documents with no prefix, in English even for other languages - To shorten vectors keep the first N values and L2-normalise again; in vLLM a dimensions request needs
--hf-overrides '{"is_matryoshka": true}', since config.json doesn't declare it - If you downloaded the Q8_0 GGUF between 15 July and 3 August 2025, fetch it again and check its SHA-256 against the Hub
- Treat retrieved documents as untrusted text before passing them to a model
Questions
Which is better for AI agents, Qwen3-Embedding-0.6B or Qwen3-Embedding-4B?
Qwen3-Embedding-4B scores 72.9 (BB) on agent readiness against Qwen3-Embedding-0.6B's 63.5 (B), and leads in 2 of 5 scored categories. Qwen3-Embedding-0.6B leads on documentation and upkeep.
Other comparisons with Qwen3-Embedding-0.6B or Qwen3-Embedding-4B
- EmbeddingGemma 300M vs Qwen3-Embedding-0.6B
- EmbeddingGemma 300M vs Qwen3-Embedding-4B
- Granite Embedding Small English R2 vs Qwen3-Embedding-0.6B
- Granite Embedding Small English R2 vs Qwen3-Embedding-4B
- jina-embeddings-v5-omni-small vs Qwen3-Embedding-0.6B
- jina-embeddings-v5-omni-small 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
- Qwen3-Embedding-4B vs Qwen3-Embedding-8B
- Qwen3-Embedding-4B vs Qwen3-VL-Embedding-2B
- Qwen3-Embedding-4B vs Qwen3-VL-Embedding-8B
Machine-readable
- This page as Markdown
/compare/qwen3-embedding-0-6b-vs-qwen3-embedding-4b.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/qwen3-embedding-0-6b.json·/api/v1/tools/qwen3-embedding-4b.json - From a terminal
anchor compare qwen3-embedding-0-6b qwen3-embedding-4b(the CLI) - Over MCP
compare_tools {"a": "qwen3-embedding-0-6b", "b": "qwen3-embedding-4b"}at/mcp, no key