Model comparison
ARK-ASR-3B vs cohere-transcribe-03-2026
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| ARK-ASR-3B | cohere-transcribe-03-2026 | |
|---|---|---|
| Parameters | 4.1B | 2.1B |
| Architecture | arkasr | cohere_asr |
| Layers | 36 | — |
| Native context | 32,768 | — |
| Mixture of experts | no | no |
| Quantizations published | 3 | 11 |
| Smallest quantization | 3.28 GiB | 1.41 GiB |
| Q4_K_M | — | 1.45 GiB |
| Licence | other | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | ARK-ASR-3B | cohere-transcribe-03-2026 | Ratio |
|---|---|---|---|
| 4,096 | 0.14 GiB | — | — |
| 8,192 | 0.28 GiB | — | — |
| 16,384 | 0.56 GiB | — | — |
| 32,768 | 1.13 GiB | — | — |
| 65,536 | 2.25 GiB | — | — |
| 131,072 | 4.50 GiB | — | — |