Model comparison
Voxtral-Mini-4B-Realtime-2602 vs cohere-transcribe-03-2026
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
| Voxtral-Mini-4B-Realtime-2602 | cohere-transcribe-03-2026 | |
|---|---|---|
| Parameters | 4.4B | 2.1B |
| Architecture | voxtral_realtime | cohere_asr |
| Layers | 26 | — |
| Native context | 131,072 | — |
| Mixture of experts | no | no |
| Quantizations published | 8 | 11 |
| Smallest quantization | 2.35 GiB | 1.41 GiB |
| Q4_K_M | 2.64 GiB | 1.45 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Voxtral-Mini-4B-Realtime-2602 | cohere-transcribe-03-2026 | Ratio |
|---|---|---|---|
| 4,096 | — | — | — |
| 8,192 | — | — | — |
| 16,384 | — | — | — |
| 32,768 | — | — | — |
| 65,536 | — | — | — |
| 131,072 | — | — | — |