MO

MOSS Transcribe preview 2B

MossGGUF by cstrUpdated 2026-08-02
Download on Hugging Face
Parameters2.42B
Context—
Layers—
Architecturecrispasr-imatrix
Smallest quant0.0 GB
Quick answer

MOSS Transcribe preview 2B needs about 3.7 GB of memory at Q4_K (a 2.8 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 0.0 GB to 4.8 GB (F16).

MOSS Transcribe preview 2B – English speech‑to‑text model

MOSS Transcribe preview 2B is an English ASR model built on a Qwen3‑1.7B‑base backbone and a Qwen3‑Omni‑MoE audio encoder. It uses a gated‑MLP adapter to project audio features into the language‑model embedding space. The model is released under Apache‑2.0.

Good at

  • Transcribing English speech
  • Low‑latency inference with CrispASR
  • Fine‑tuned on Open ASR Leaderboard

Limits

  • Designed for English only
  • Performance drops on heavy accents or noisy audio
  • Requires Whisper log‑mel frontend

Best for

  • Research on English ASR
  • Transcription of 16 kHz mono audio
  • Evaluation on Open ASR Leaderboard

Who it’s for. Developers and researchers needing an open‑source English speech‑to‑text model. It is suitable for integration into pipelines that accept 16 kHz mono audio and use the CrispASR runtime.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 0.5 GB
Fits Get file →
Q3_K2.6 GB 3.4 GB
Fits Get file →
Q4_K2.8 GB 3.7 GB
Fits Get file →
Q4_K2.8 GB 3.7 GB
Fits Get file →
Q8_03.5 GB 4.4 GB
Fits Get file →
F164.8 GB 5.9 GB
Fits Get file →

For 8 GB VRAM, the Q4_K build (file size 2.82 GB, estimated 3.7 GB memory at 8k context) is recommended. If more VRAM is available, the Q8_0 build (3.52 GB, 4.4 GB memory) offers higher precision. The F16 build (4.84 GB, 5.9 GB memory) requires the largest memory budget.

What GPU do you need for MOSS Transcribe preview 2B?

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 (2.8 GB file, about 3.7 GB in use).
  • 6 GB VRAM: the best quant that fits is F16 (4.8 GB file, about 5.9 GB in use).

Which MOSS Transcribe preview 2B quant should you choose?

Q4_K (2.8 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.

If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturecrispasr-imatrix
Parameters2.42B

Source

Quantized bycstr
FamilyMoss
Licenseapache-2.0
Downloads (HF)744
Last updated2026-08-02

Frequently asked questions about MOSS Transcribe preview 2B

How much VRAM does MOSS Transcribe preview 2B need?

About 3.7 GB at Q4_K with an 8K context window. That is the 2.8 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 MOSS Transcribe preview 2B on an 8 GB GPU?

Yes, with F16: the 4.8 GB file needs about 5.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of MOSS Transcribe preview 2B GGUF?

It depends on the quant: file is 0.0 GB, Q3_K is 2.6 GB, Q4_K is 2.8 GB, Q4_K is 2.8 GB, Q8_0 is 3.5 GB, F16 is 4.8 GB.

Where can I download MOSS Transcribe preview 2B GGUF?

The files are on Hugging Face in the cstr/MOSS-Transcribe-preview-2B-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 MOSS Transcribe preview 2B 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 MOSS Transcribe preview 2B on 8 GB VRAM?

Yes, the Q4_K quantization uses about 3.7 GB of VRAM at 8k context, which fits within an 8 GB GPU.

Does the model support non‑English speech?

No, it is designed for English ASR and may perform poorly on other languages.

What audio format is required?

Audio must be 16 kHz mono with Whisper log‑mel features (128 mel bins, 400 FFT size, 160 hop length).

Other GGUFs of MOSS-Transcribe-preview-2B

Similar models

Browse all →

More from cstr

All by cstr →

More Moss models

All Moss →

Other tiny models

All tiny models →