parakeet-tdt-0.6b-v2
nvidia/parakeet-tdt-0.6b-v2parakeet-tdt-0.6b-v2 at Q4_K_M is exactly 475,491,840 bytes (0.44 GiB / 0.48 GB) — an effective 6.157 bits per weight, not the nominal 4.
Shipped quantizations
| Quant | Size● | Exact bytes● | Effective bpw● | Tensors● | Publisher |
|---|---|---|---|---|---|
| Q4_K | 0.37 GiB | 396,937,984 | 5.139 | — | cstr |
| Q4_K_M | 0.44 GiB | 475,491,840 | 6.157 | 697 | handy-computer |
| Q5_K_M | 0.50 GiB | 539,012,608 | 6.979 | 697 | handy-computer |
| Q6_K | 0.56 GiB | 600,408,576 | 7.774 | 697 | handy-computer |
| Q8_0 | 0.62 GiB | 664,466,368 | 8.603 | 725 | cstr |
| Q8_0 | 0.68 GiB | 729,574,912 | 9.446 | 697 | handy-computer |
| F16 | 1.15 GiB | 1,237,334,592 | 16.021 | 697 | handy-computer |
| F32 | 2.30 GiB | 2,471,550,272 | 32.001 | — | handy-computer |
Measured
| Metric | Value | What it means |
|---|---|---|
| rtf | 6038.1 | |
| RTFx | 6038.1 | Higher is better — audio seconds processed per second of compute. |
| Word error rate | 2.73% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 10.40% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 1.27% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 8.15% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 10.78% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 9.10% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 5.68% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 1.94% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 3.78% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 8.23% | Lower is better — the share of words transcribed incorrectly. |
| Word error rate | 5.39% | Lower is better — the share of words transcribed incorrectly. |
RTFx measured by the Open ASR Leaderboard on a single datacenter GPU at a large batch size. It ranks models against each other; it says nothing about throughput on consumer hardware. We reproduce these figures with attribution; they are not ours and we have not verified the runs. Source: open-asr-leaderboard-english-short-latest.
Compare with
Will it run on your card?
Why other calculators give a different number
A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 0.32 GiB. The real file is 0.44 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.
Architecture
Architecture unavailable — this repository is gated and no ungated mirror was found. Exact file sizes above are still authoritative; only the KV math needs the config.
Questions people ask
- How much VRAM does parakeet-tdt-0.6b-v2 need?
- Q4_K_M is exactly 475,491,840 bytes (0.44 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
- Which quantization of parakeet-tdt-0.6b-v2 should I use?
- Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.