Model comparison

canary-qwen-2.5b vs Voxtral-Mini-4B-Realtime-2602

At Q4_K_M, canary-qwen-2.5b is the smaller download — 1,737,575,808 bytes against 2,830,493,984.

From the file· summed bytes, KV per layer

Side by side

canary-qwen-2.5bVoxtral-Mini-4B-Realtime-2602
Parameters2.6B4.4B
Architecturecanary_qwenvoxtral_realtime
Layers26
Native context131,072
Mixture of expertsnono
Quantizations published98
Smallest quantization1.62 GiB2.35 GiB
Q4_K_M1.62 GiB2.64 GiB
Licencecc-by-4.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextcanary-qwen-2.5bVoxtral-Mini-4B-Realtime-2602Ratio
4,096
8,192
16,384
32,768
65,536
131,072