KaLM Embedding Gemma3 12B 2511 vs Qwen 3 Embedding 8B

Wins 14 of 133 benchmarks
$0.01 / 1M tokens
Wins 18 of 133 benchmarks
KaLM Embedding Gemma3 12B 251172.3
Qwen 3 Embedding 8B70.6
KaLM Embedding Gemma3 12B 25115.5
Qwen 3 Embedding 8B10.1
KaLM Embedding Gemma3 12B 25115.5
Qwen 3 Embedding 8B10.1
KaLM Embedding Gemma3 12B 25115.5
Qwen 3 Embedding 8B10.1
KaLM Embedding Gemma3 12B 251192.8
Qwen 3 Embedding 8B93
KaLM Embedding Gemma3 12B 251150.7
Qwen 3 Embedding 8B50.4
KaLM Embedding Gemma3 12B 25115.5
Qwen 3 Embedding 8B10.1
KaLM Embedding Gemma3 12B 251125.2
Qwen 3 Embedding 8B23.7
KaLM Embedding Gemma3 12B 251163.1
Qwen 3 Embedding 8B62.6
KaLM Embedding Gemma3 12B 251169.2
Qwen 3 Embedding 8B79.2
KaLM Embedding Gemma3 12B 251152.9
Qwen 3 Embedding 8B55.7
KaLM Embedding Gemma3 12B 251182.5
Qwen 3 Embedding 8B86.6
KaLM Embedding Gemma3 12B 251187.6
Qwen 3 Embedding 8B86.8
KaLM Embedding Gemma3 12B 251163.4
Qwen 3 Embedding 8B61.6
KaLM Embedding Gemma3 12B 251162.5
Qwen 3 Embedding 8B61.7
KaLM Embedding Gemma3 12B 251183.8
Qwen 3 Embedding 8B80.9
KaLM Embedding Gemma3 12B 251177.9
Qwen 3 Embedding 8B74
KaLM Embedding Gemma3 12B 251155.8
Qwen 3 Embedding 8B57.6
KaLM Embedding Gemma3 12B 25115.5
Qwen 3 Embedding 8B10.1
KaLM Embedding Gemma3 12B 251133
Qwen 3 Embedding 8B28.7
KaLM Embedding Gemma3 12B 251184.7
Qwen 3 Embedding 8B86.4
KaLM Embedding Gemma3 12B 251167.3
Qwen 3 Embedding 8B65.6
KaLM Embedding Gemma3 12B 251175.7
Qwen 3 Embedding 8B70.9
KaLM Embedding Gemma3 12B 251179
Qwen 3 Embedding 8B81.1
KaLM Embedding Gemma3 12B 251190.5
Qwen 3 Embedding 8B91.2
KaLM Embedding Gemma3 12B 251131.6
Qwen 3 Embedding 8B22
KaLM Embedding Gemma3 12B 251161.6
Qwen 3 Embedding 8B71.3
KaLM Embedding Gemma3 12B 251182.7
Qwen 3 Embedding 8B84.7
KaLM Embedding Gemma3 12B 251177.7
Qwen 3 Embedding 8B84.4
KaLM Embedding Gemma3 12B 251176.2
Qwen 3 Embedding 8B67.7
KaLM Embedding Gemma3 12B 251157.2
Qwen 3 Embedding 8B58.2
KaLM Embedding Gemma3 12B 251185.1
Qwen 3 Embedding 8B88.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B62.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B62.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B62.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B0.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B50.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B63.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B74.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B79.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B79.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B79.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B59
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B59
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B54.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B63.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B56.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B87.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B73.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B75
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B77.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B80.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B84.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B67
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B78.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B63.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B80.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B80.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B80.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B64.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B79.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B69.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B61.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B74.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B93.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B66.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B81.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B62.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B68.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B83.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B89
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B71.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B86.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B30.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B73.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B94.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B77.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B75.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B69.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B70.9
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B77.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B67.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B78.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B89.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B45.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B86.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B65.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B70.1
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B73
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B49
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B41.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B82.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B66.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B87
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B73.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B71.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B75.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B50.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B87.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B72.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B70.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B86
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B71.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B69.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B74.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B66.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B44.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B70.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B78.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B79.3
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B70.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B71.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B74.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B55.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B71.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B92.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B82.5
KaLM Embedding Gemma3 12B 251166.4
Qwen 3 Embedding 8B—
KaLM Embedding Gemma3 12B 251166.8
Qwen 3 Embedding 8B—
KaLM Embedding Gemma3 12B 251151.7
Qwen 3 Embedding 8B—
KaLM Embedding Gemma3 12B 251163.7
Qwen 3 Embedding 8B—
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B66.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B72.7
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B59.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B67
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B67.2
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B34.8
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B58.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B90.4
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B51.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B88.6
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B87.5
KaLM Embedding Gemma3 12B 2511—
Qwen 3 Embedding 8B69.4

Specs

SpecKaLM Embedding Gemma3 12B 2511Qwen 3 Embedding 8B
Price—!$0.01
Context33K33K
WeightsOpenOpen
ReleasedNov 2025Jun 2025
Published speed—3.3 × vLLM H100 (Baseten)
Dimensions3,8404,096
Max input131K tokens33K tokens
Memory (4-bit)7.1 GB4.5 GB
Licencetencent-kalm-embedding-communityApache-2.0
Commercial useRestrictedAllowed

What breaks when you switch

KaLM Embedding Gemma3 12B 2511 → Qwen 3 Embedding 8B
Re-embed your whole corpus›

Vectors from different models can't be mixed; dimensions change from 3,840 to 4,096.

Max input: 131K tokens → 33K tokens
Qwen 3 Embedding 8B → KaLM Embedding Gemma3 12B 2511
Re-embed your whole corpus›

Vectors from different models can't be mixed; dimensions change from 4,096 to 3,840.

Memory (4-bit): 4.5 GB → 7.1 GB