GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
NAMAA-Space
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
NAMAA-Space
Model Tree
Base
CohereLabs/cohere-transcribe-arabic-07-2026
Quantized
this model
Input Modalities
Audio
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
| Base model | CohereLabs/cohere-transcribe-arabic-07-2026 |
| Task | Automatic speech recognition (audio → text) |
| Architecture | Conformer encoder + Transformer decoder (attention encoder–decoder) |
| Languages | Arabic (incl. dialects), English |
| Quantization | bitsandbytes NF4 4-bit with double quantization |
| Disk size | 1.47 GB (vs. 4.13 GB fp16) |
| License | Apache 2.0 |
Word Error Rate of this quantized model measured against the full fp16 model on 1,009 Saudi-dialect podcast clips (Arabic text normalized):
| Model | Size | WER vs. full fp16 |
|---|---|---|
| Full (fp16) | 4.13 GB | — (reference) |
| This model (INT4 (NF4)) | 1.47 GB | 1.89% |
The int8 build is identical to full-precision in WER; the int4 build differs by ~1.9%, mostly minor orthographic variation (e.g. hamza placement).
python
import torchfrom transformers import AutoProcessor, CohereAsrForConditionalGenerationfrom transformers.audio_utils import load_audioprocessor = AutoProcessor.from_pretrained("NAMAA-Space/cohere-transcribe-arabic-07-2026-int4")model = CohereAsrForConditionalGeneration.from_pretrained("NAMAA-Space/cohere-transcribe-arabic-07-2026-int4", device_map="auto")audio = load_audio("your_audio.wav", sampling_rate=16000)inputs = processor(audio, sampling_rate=16000, language="ar", return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=256)print(processor.decode(outputs, skip_special_tokens=True))
Requirements: transformers, bitsandbytes, soundfile/librosa, and a CUDA GPU.
python
import torchfrom transformers import BitsAndBytesConfigquantization_config = BitsAndBytesConfig(load_in_4bit=True,bnb_4bit_quant_type="nf4",bnb_4bit_compute_dtype=torch.bfloat16,bnb_4bit_use_double_quant=True,)
| # | Full (fp16) | This model |
|---|---|---|
| 1 | السلام عليكم ورحمة الله وبركاته. مساكم الله بالخير وحياكم الله معنا في ثالث حلقات لقا… | السلام عليكم ورحمة الله وبركاته. مساكم الله بالخير وحياكم الله معنا في ثالث حلقات لقا… |
| 2 | اهتمامات ولكن ولكن هذا لا يرقى انا عندي احصائيات قديمه وذكرت هذا في اكثر من مره في ال… | اهتمامات ولكن ولكن هذا لا يرقى انا عندي احصائيات قديمه وذكرت هذا في اكثر من مره في ال… |
| 3 | يتبع اوامر الله سبحانه وتعالى، ويصدق اخباره، ويرضى باقداره. هذا هو الذي يحسن الادب مع… | يتبع اوامر الله سبحانه وتعالى، ويصدق اخباره، ويرضى باقداره. هذا هو الذي يحسن الادب مع… |
transformers + bitsandbytes and requires a CUDA GPU. It is
not directly servable with vLLM (use a GPTQ / AWQ / compressed-tensors build for vLLM).Released under Apache 2.0. This is a quantized derivative of
CohereLabs/cohere-transcribe-arabic-07-2026 by Cohere and Cohere Labs; all original terms apply.