What We Know
The Qwen3-Embedding series is a family of text embedding and reranking models released by the Alibaba Qwen team in June 2025. It comes in three sizes (0.6B, 4B, and 8B), and here I summarize the performance of the smallest one, the 0.6B version.
Basic Specs
| Item | Value |
|---|---|
| Parameters | About 600 million (0.6B), 28 transformer layers |
| Context length | 32K tokens |
| Embedding dimension | Up to 1024 (user-defined between 32 and 1024, MRL) |
| Supported languages | 100+ (natural languages + programming languages) |
| Instruction awareness | Supported (task-specific instructions can be provided) |
Benchmark Performance
MTEB Multilingual (MMTEB)
| Model | Size | Average score |
|---|---|---|
| Qwen3-Embedding-0.6B | 0.6B | 64.33 |
| multilingual-e5-large-instruct | 0.6B | 63.22 |
| Cohere embed-multilingual-v3.0 | - | 61.12 |
| Gemini Embedding | - | 68.37 |
| Qwen3-Embedding-4B | 4B | 69.45 |
| Qwen3-Embedding-8B (#1 on the leaderboard, 2025-06) | 8B | 70.58 |
The 0.6B model edges out a leading open-source model of similar size (multilingual-e5-large-instruct) and also surpasses Cohere's commercial multilingual embeddings.
Comparison with OpenAI (text-embedding-3)
| Item | Qwen3-Embedding-0.6B | text-embedding-3-small | text-embedding-3-large |
|---|---|---|---|
| MTEB (English) | 70.70 (v2) | 62.3 | 64.6 |
| Multilingual benchmark (MTEB/MIRACL) | 64.33 (MMTEB) | 44.0 (MIRACL) | 54.9 (MIRACL) |
| Max context | 32K | 8K | 8K |
| Embedding dimension | 32~1024 (MRL) | 512 / 1536 | 256 / 1024 / 3072 |
| Price | Free (local) | $0.02 / 1M tokens | $0.13 / 1M tokens |
| Availability | Open weights (Apache 2.0) | Closed API | Closed API |
Note: OpenAI's official scores are based on the older MTEB/MIRACL, while the Qwen3 numbers are based on MMTEB v2, so the axes are not exactly the same. Still, the finding that the 0.6B open model outperforms
text-embedding-3-largeon both English retrieval and multilingual tasks is reproduced across several independent leaderboards.