NE

NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16

NemotronGGUF by JmicaUpdated 2026-10-31
Download on Hugging Face
Parameters38B
Context1024K
Layers53
Architecturenemotron_h_moe
Smallest quant13.9 GB
Quick answer

NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 needs about 24.9 GB of memory at Q3_K_S (a 17.3 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 4 quants, from 13.9 GB (Q2_K) to 21.7 GB (Q4_K_S). Longer context needs more: at 32K tokens the KV cache alone adds about 28.5 GB.

NVIDIA Nemotron 3.5 Transcribe 30B-A4B – offline speech and audio model

NVIDIA Nemotron 3.5 Transcribe is an offline speech and audio model that transcribes recordings up to one hour and supports a 65,536-token context. It uses a hybrid mixture-of-experts architecture with approximately 4.3B active parameters. The model is available in BF16, FP8, and NVFP4 precision.

Good at

  • Multilingual transcription
  • Speaker‑attributed transcription
  • Audio question answering

Limits

  • No vision support
  • Requires large VRAM for full precision
  • License is other

Best for

  • Transcribing long audio files
  • Generating timestamps and speaker labels
  • Answering audio‑based questions

Who it’s for. Developers building offline transcription services or applications that need multilingual audio processing. The model is suitable for both commercial and non‑commercial use.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K13.9 GB 21.6 GB
Too big Get file →
Q3_K_S17.3 GB 24.9 GB
Too big Get file →
IQ4_NL20.4 GB 28.0 GB
Too big Get file →
Q4_K_S21.7 GB 29.3 GB
Too big Get file →

For 8 GB VRAM use Q2_K (file size 13.95 GB, estimated 21.6 GB memory at 8 k context). For 12 GB VRAM use Q3_K_S (file size 17.29 GB, estimated 24.9 GB memory). For 16 GB VRAM use IQ4_NL (file size 20.37 GB, estimated 28 GB memory). For 20 GB VRAM use Q4_K_S (file size 21.67 GB, estimated 29.3 GB memory).

What GPU do you need for NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: the best quant that fits is Q2_K (13.9 GB file, about 21.6 GB in use).
  • 32 GB VRAM: the best quant that fits is Q4_K_S (21.7 GB file, about 29.3 GB in use).

Which NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 quant should you choose?

Q3_K_S (17.3 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

If memory is tight, the smallest file is Q2_K at 13.9 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturenemotron_h_moe
Parameters38B (approximate, mixture of experts)
Active parameters4B
Layers53
Trained context1,048,576 tokens
KV cache (fp16)848 KB per token

Source

Quantized byJmica
FamilyNemotron
Licenseother
Downloads (HF)1,239
Last updated2026-10-31

Frequently asked questions about NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16

How much VRAM does NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 need?

About 24.9 GB at Q3_K_S with an 8K context window. That is the 17.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 21.6 GB, so you would need a larger GPU or CPU offloading.

What is the file size of NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 GGUF?

It depends on the quant: Q2_K is 13.9 GB, Q3_K_S is 17.3 GB, IQ4_NL is 20.4 GB, Q4_K_S is 21.7 GB.

What context length does NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 support?

The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 7.1 GB at 8K and 28.5 GB at 32K.

Where can I download NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 GGUF?

The files are on Hugging Face in the Jmica/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does NVIDIA-Nemotron-3.5-Transcribe-30B-A4B BF16 use?

Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run NVIDIA Nemotron 3.5 Transcribe on 8 GB VRAM?

Yes, using the Q2_K quantized model which requires about 21.6 GB of memory at 8 k context, but it will need more VRAM than 8 GB, so it is not feasible on 8 GB alone.

Does this model support vision tasks?

No, the model only supports audio and text tasks.

What precision formats are available for this model?

The model is released in BF16, FP8, and NVFP4 precision.

Other GGUFs of NVIDIA-Nemotron-3.5-Transcribe-30B-A4B

Similar models

Browse all →

More from Jmica

All by Jmica →

More Nemotron models

All Nemotron →

Other large models

All large models →