Model comparison
Qwen3-Embedding-0.6B vs KaLM-embedding-multilingual-mini-instruct-v2.5
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: KaLM-embedding-multilingual-mini-instruct-v2.5's KV cache at 32K is 9.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3-Embedding-0.6B | KaLM-embedding-multilingual-mini-instruct-v2.5 | |
|---|---|---|
| Parameters | 596M | 494M |
| Architecture | qwen3 | qwen2 |
| Layers | 28 | 24 |
| Native context | 32,768 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 22 | 1 |
| Smallest quantization | 0.28 GiB | 0.49 GiB |
| Q4_K_M | 0.37 GiB | — |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-Embedding-0.6B | KaLM-embedding-multilingual-mini-instruct-v2.5 | Ratio |
|---|---|---|---|
| 4,096 | 0.44 GiB | 0.05 GiB | 9.33× |
| 8,192 | 0.88 GiB | 0.09 GiB | 9.33× |
| 16,384 | 1.75 GiB | 0.19 GiB | 9.33× |
| 32,768 | 3.50 GiB | 0.38 GiB | 9.33× |
| 65,536 | 7.00 GiB | 0.75 GiB | 9.33× |
| 131,072 | 14.00 GiB | 1.50 GiB | 9.33× |