Harrier OSS v1 27B vs KaLM Embedding Gemma3 12B 2511
MTEB multilingual v2 · Mean over tasks · multilingual (1037 languages) Embeddings assign examples to categories, group similar examples, estimate semantic similarity between pairs, classify relationships between pairs, rank relevant documents for queries, find translated sentence pairs, rerank documents according to instructions, assign multiple labels to each example, reorder candidate documents by relevance on MTEB Multilingual v2 · mean over tasks in 1037 languages including 'Auhelawa, Abau, Achi, Achinese, Achuar-Shiwiar, Adzera, Afrikaans, Agarabi; tasks receive equal weight
Harrier OSS v1 27B74.3
KaLM Embedding Gemma3 12B 251172.3
| Benchmark | Harrier OSS v1 27B | KaLM Embedding Gemma3 12B 2511 |
|---|---|---|
| MTEB multilingual v2 · Mean over tasks · multilingual (1037 languages) | 74.3 | 72.3 |
Specs
| Spec | Harrier OSS v1 27B | KaLM Embedding Gemma3 12B 2511 |
|---|---|---|
| Context | 33K | 33K |
| Weights | Open | Open |
| Released | Mar 2026 | Nov 2025 |
| Dimensions | 5,376 | 3,840 |
| Max input | 33K tokens | 131K tokens |
| Languages | en, Esperanto, es +90 | — |
| Memory (4-bit) | 16 GB | 7.1 GB |
| Licence | MIT | tencent-kalm-embedding-community |
| Commercial use | Allowed | Restricted |
What breaks when you switch
Harrier OSS v1 27B → KaLM Embedding Gemma3 12B 2511
Re-embed your whole corpus›
Vectors from different models can't be mixed; dimensions change from 5,376 to 3,840.
Languages: not verified
KaLM Embedding Gemma3 12B 2511 → Harrier OSS v1 27B
Re-embed your whole corpus›
Vectors from different models can't be mixed; dimensions change from 3,840 to 5,376.