Model comparison

cohere-transcribe-03-2026 vs Voxtral-Mini-4B-Realtime-2602

At Q4_K_M, cohere-transcribe-03-2026 is the smaller download — 1,558,162,944 bytes against 2,830,493,984.

From the file· summed bytes, KV per layer

Side by side

cohere-transcribe-03-2026Voxtral-Mini-4B-Realtime-2602
Parameters2.1B4.4B
Architecturecohere_asrvoxtral_realtime
Layers26
Native context131,072
Mixture of expertsnono
Quantizations published118
Smallest quantization1.41 GiB2.35 GiB
Q4_K_M1.45 GiB2.64 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextcohere-transcribe-03-2026Voxtral-Mini-4B-Realtime-2602Ratio
4,096
8,192
16,384
32,768
65,536
131,072