What GPU do I need to run salesforce-sovereign-intelligence/SOV-Transcribe-1?
26.6B parameters, published in BF16. View on Hugging Face
SOV-Transcribe-1 is published by salesforce-sovereign-intelligence on Hugging Face, with 463 downloads and 58 likes to date. It's a SovereignAITranscribeForConditionalGeneration model built for automatic-speech-recognition, published natively in BF16.
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision, plus a fixed overhead for activations and allocator fragmentation. Speech models don't build the same growing KV-cache a text model does. Memory scales primarily with input audio length. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 49.6 GB | 59.5 GB | A100 | 1 | $1.31/hr |
| cheaper alt. | RTX A5000 | 3 | $0.528/hr | ||
| FP8 (quantized) | 24.8 GB | 29.8 GB | RTX 5000 Ada | 1 | $0.539/hr |
| cheaper alt. | RTX 5060 Ti | 2 | $0.290/hr | ||
| INT4 (quantized) | 12.4 GB | 14.9 GB | RTX 5060 Ti | 1 | $0.145/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run SOV-Transcribe-1 at its published (BF16) precision: 1× A100, at $1.31/hr per GPU ($1.31/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
SOV-Transcribe-1: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Sliding-window attention: windowed layers stop growing at the window, the full-attention layers keep growing. Cached per token: 245,760 bytes at 16-bit while the context is inside the window.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.35 GB |
| 32K | 1.4 GB |
| 128K | 5.2 GB |
| 256K (model maximum) | 10.2 GB |
Computed from the layer, head and window counts in the model's own configuration (salesforce-sovereign-intelligence/SOV-Transcribe-1), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
SOV-Transcribe-1: common questions
Can SOV-Transcribe-1 run on a single GPU?
Yes, but not on a desktop card. At BF16 it needs 59.5 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest live single-card fit is a 80.0 GB A100 at $1.31/hr.
What is the least VRAM SOV-Transcribe-1 can run in?
14.9 GB, at INT4 (quantized), which fits a 16 GB card, against 59.5 GB at BF16. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.
Does quantizing SOV-Transcribe-1 lower the GPU bill?
Yes. At BF16 the cheapest live fit is one A100 at $1.31/hr. At INT4 (quantized) it drops to one RTX 5060 Ti at $0.145/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More Gemma 4 models
- gemma-4-26B-A4B-it-FP8-dynamic (26.5B, F8_E4M3)
- gemma-4-26B-A4B-it-qat-q4_0-unquantized (26.5B, BF16)
- gemma-4-26B-A4B-it (25.8B, BF16)
- GEV-26B-Decide (25.8B, BF16)
- rune-26b-a4b-GGUF (25.8B, BF16)
Lineage
Based on gemma-4-26B-A4B-it.
Alternatives at this size
Other models for automatic-speech-recognition within about a third of SOV-Transcribe-1's 26.6B parameters, from other model lines.
- NVIDIA-Nemotron-3.5-Transcribe-30B-A4B (33.7B, BF16)
More on SOV-Transcribe-1
Fits on an 80 GB GPU at BF16: every model that fits in 80 GB.
Fits on a 32 GB GPU at FP8: every model that fits in 32 GB.
Fits on a 16 GB GPU at INT4: every model that fits in 16 GB.
Related reading: A100 pricing and specs, Whisper variants compared, What AI inference is, and What AI model fine-tuning is.
Your own GPU cloud accounts
One console for the GPU clouds you already rent from. Your keys stay on your machine.
$20 a month, flat