GR

granite speech 4.0 1b

GraniteGGUF by cstrUpdated 2026-08-02
Download on Hugging Face
Parameters2.31B
Context—
Layers—
Architecturegranite_speech
Smallest quant2.9 GB
Quick answer

granite speech 4.0 1b needs about 4.8 GB of memory at Q8_0 (a 3.9 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 7 quants, from 2.9 GB (Q4_K) to 5.6 GB (F16).

Granite Speech 4.0 1B – 2.31B parameter speech‑LLM

Granite Speech 4.0 1B is a 2.31B parameter speech‑LLM based on a Conformer encoder, BLIP‑2 Q‑Former projector, and Granite‑4.0 1B LLM head. It supports English, French, German, Spanish, Portuguese, and Japanese. The GGUF conversion includes full GPT‑2‑style BPE vocab and merges for prompt‑based translation.

Good at

  • Speech transcription
  • Speech translation
  • Prompt‑based audio‑to‑text

Limits

  • No vision support
  • Requires 8k context for estimates
  • License is apache‑2.0

Best for

  • Embedded speech assistants
  • Low‑latency transcription on edge devices
  • Batch audio processing

Who it’s for. Developers building speech‑centric applications who need a lightweight, fully C++ runtime model. The model is ideal for projects that require prompt‑based translation and do not need vision capabilities.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K2.9 GB 3.8 GB
Fits Get file →
Q4_K2.9 GB 3.8 GB
Fits Get file →
Q5_03.2 GB 4.1 GB
Fits Get file →
Q8_03.9 GB 4.8 GB
Fits Get file →
Q8_03.9 GB 4.8 GB
Fits Get file →
—5.6 GB 6.7 GB
Fits Get file →
F165.6 GB 6.8 GB
Fits Get file →

For 8 GB VRAM, use the Q4_K quant (file size 2.94 GB, estimated memory 3.8 GB). For 12 GB VRAM, Q5_0 (3.17 GB, 4.1 GB). For 16 GB VRAM, Q8_0 (3.86 GB, 4.8 GB). The F16 quant (5.58 GB, 6.8 GB) is suitable for 24 GB VRAM or more.

What GPU do you need for granite speech 4.0 1b?

These figures add the file size, an fp16 KV cache at an 8K context window 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: the best quant that fits is Q4_K (2.9 GB file, about 3.8 GB in use).
  • 6 GB VRAM: the best quant that fits is Q8_0 (3.9 GB file, about 4.8 GB in use).
  • 8 GB VRAM: the best quant that fits is F16 (5.6 GB file, about 6.8 GB in use).

Which granite speech 4.0 1b quant should you choose?

Q8_0 (3.9 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.

If memory is tight, the smallest file is Q4_K at 2.9 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegranite_speech
Parameters2.31B

Source

Quantized bycstr
FamilyGranite
Licenseapache-2.0
Downloads (HF)919
Last updated2026-08-02

Frequently asked questions about granite speech 4.0 1b

How much VRAM does granite speech 4.0 1b need?

About 4.8 GB at Q8_0 with an 8K context window. That is the 3.9 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 granite speech 4.0 1b on an 8 GB GPU?

Yes, with F16: the 5.6 GB file needs about 6.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of granite speech 4.0 1b GGUF?

It depends on the quant: Q4_K is 2.9 GB, Q4_K is 2.9 GB, Q5_0 is 3.2 GB, Q8_0 is 3.9 GB, Q8_0 is 3.9 GB, file is 5.6 GB and 1 more.

Where can I download granite speech 4.0 1b GGUF?

The files are on Hugging Face in the cstr/granite-speech-4.0-1b-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 granite speech 4.0 1b use?

Hugging Face lists the license as apache-2.0. 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 Granite Speech 4.0 1B on 8 GB VRAM?

Yes, the Q4_K quant uses 3.8 GB estimated memory at 8k context, fitting within 8 GB VRAM.

Does this model support vision tasks?

No, it is a speech‑only model with no vision support.

What languages can it transcribe?

English, French, German, Spanish, Portuguese, and Japanese.

Similar models

Browse all →

More from cstr

All by cstr →

More Granite models

All Granite →

Other tiny models

All tiny models →