cohere-transcribe-03-2026

Repository
CohereLabs/cohere-transcribe-03-2026
Engine
vLLM
Listed
11 Aug 2026

Cohere Transcribe is an open source release of a 2B parameter dedicated audio-in, text-out automatic speech recognition (ASR) model. The model supports 14 languages.

Serves
Chat, Transcription
Parameters
2.1B
Context
1,024 tokens
Weights
4.1 GB
License
Apache-2.0

Capabilities

Read from the model's capability tags. Requests are admitted per tag.

Chat chat
  • POST /v1/chat/completions
  • POST /v1/responses

Endpoint task mode generate

Transcription transcription
  • POST /v1/audio/transcriptions
  • POST /v1/audio/translations

Endpoint task mode transcribe

Inputs
Text
Reasoning effort reasoning_effort
does not reason
Structured output structured_output
not yet probed
Admitted until the capability probe observes the model.
Forced tool choice forced_tool_choice
not yet probed
Admitted until the capability probe observes the model.

Launch settings

Every engine serving the model starts with these.

Set by the owner. auto: the serving node derives it. default: the platform's value, left in place.

Engine
vLLM
Context window --max-model-len
1,024 tokens
Quantization --quantization
none default
Weights load as stored.
KV cache dtype --kv-cache-dtype
auto
No flag: the checkpoint's KV quantization, else the model dtype.
Scheduling policy --scheduling-policy
priority default
Interactive requests ahead of batch work; same-class traffic in arrival order.
Tool-call parser --tool-call-parser
auto
Resolves to pythonic from the architecture.
Reasoning parser --reasoning-parser
auto
No parser for this architecture.
Trust remote code --trust-remote-code
on default
Python shipped in the bundle may run at load (the default).

Engine tuning

Every setting auto: the serving node derives each one.

Set by the owner. auto: the serving node derives it.

Mixture of experts

Expert parallel --enable-expert-parallel
auto
Expert load balancing --enable-eplb
auto
All-to-all backend --all2all-backend
auto

Memory and batching

GPU memory utilization --gpu-memory-utilization
auto
Max sequences --max-num-seqs
auto
Max batched tokens --max-num-batched-tokens
auto
Long-prefill threshold --long-prefill-token-threshold
auto

Execution

Enforce eager --enforce-eager
auto
Prefix caching --enable-prefix-caching
auto
Async scheduling --no-async-scheduling
auto
Custom all-reduce --disable-custom-all-reduce
auto

Speculative decoding

Draft tokens checked by the model in one pass.

Speculative decoding
not configured
An endpoint's own setting, or the serving node's default for the model family, applies.

Architecture

Read from the model's files at upload.

Architecture
CohereAsrForConditionalGeneration
Parameters
2,065,931,488
Vocabulary vocab_size
16,384
Weights on disk
4.1 GB, safetensors

Call it

Create an endpoint for the model, then send it a request in the OpenAI shape.

Shell
# Create an endpoint for this model, then provision it. # --tier-id is the service tier to run on; pick one at /endpoints/new. erectl endpoints create \ --name "My endpoint" \ --slug my-endpoint \ --model-id 21882E3F-BECF-4E0C-A748-C74F8C921E38 \ --tier-id <tier-id> # provision takes the endpoint UUID returned by create erectl endpoints provision <endpoint-id> # Call it. Requests are routed by /<project-id>/<endpoint-slug>. curl https://api.erebine.ai/proj_ABC123/my-endpoint/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $EREBINE_API_KEY" \ -d '{ "model": "cohere-transcribe-03-2026", "messages": [{"role": "user", "content": "Hello, world!"}], "stream": true }'

Model card

As published with CohereLabs/cohere-transcribe-03-2026. Images are left out; its HTML tables are set as tables.

Cohere Transcribe

Developed by: Cohere and Cohere Labs. Point of Contact: Cohere Labs.

Namecohere-transcribe-03-2026
Architectureconformer-based encoder-decoder
Inputaudio waveform β†’ log-Mel spectrogram. Audio is automatically resampled to 16kHz if necessary during preprocessing. Similarly, multi-channel (stereo) inputs are averaged to produce a single channel signal.
Outputtranscribed text
Model size2B
Modela large Conformer encoder extracts acoustic representations, followed by a lightweight Transformer decoder for token generation
Training objectivesupervised cross-entropy on output tokens; trained from scratch
LanguagesTrained on 14 languages: European: English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish APAC: Chinese (Mandarin), Japanese, Korean, Vietnamese MENA: Arabic
LicenseApache 2.0

✨Try the Cohere Transcribe demo✨

Usage

Cohere Transcribe is supported natively in transformers. This is the recommended way to use the model for
offline inference. For online inference, see the vLLM integration example below.

pip install transformers>=5.4.0 torch huggingface_hub soundfile librosa sentencepiece protobuf accelerate
pip install datasets  # only needed for long-form and non-English examples

Testing was carried out with torch==2.10.0 but it is expected to work with other versions.

Quick Start πŸ€—

Transcribe any audio file in a few lines:

from transformers import AutoProcessor, CohereAsrForConditionalGeneration
from transformers.audio_utils import load_audio
from huggingface_hub import hf_hub_download

processor = AutoProcessor.from_pretrained("CohereLabs/cohere-transcribe-03-2026")
model = CohereAsrForConditionalGeneration.from_pretrained("CohereLabs/cohere-transcribe-03-2026", device_map="auto")

audio_file = hf_hub_download(
    repo_id="CohereLabs/cohere-transcribe-03-2026",
    filename="demo/voxpopuli_test_en_demo.wav",
)
audio = load_audio(audio_file, sampling_rate=16000)

inputs = processor(audio, sampling_rate=16000, return_tensors="pt", language="en")
inputs.to(model.device, dtype=model.dtype)

outputs = model.generate(**inputs, max_new_tokens=256)
text = processor.decode(outputs, skip_special_tokens=True)
print(text)

Long-form transcription

For audio longer than the feature extractor's max_audio_clip_s, the feature extractor automatically splits the waveform into chunks.
The processor reassembles the per-chunk transcriptions using the returned audio_chunk_index.

This example transcribes a 55 minute earnings call:

from transformers import AutoProcessor, CohereAsrForConditionalGeneration
from datasets import load_dataset
import time

processor = AutoProcessor.from_pretrained("CohereLabs/cohere-transcribe-03-2026")
model = CohereAsrForConditionalGeneration.from_pretrained("CohereLabs/cohere-transcribe-03-2026", device_map="auto")

ds = load_dataset("distil-whisper/earnings22", "full", split="test", streaming=True)
sample = next(iter(ds))

audio_array = sample["audio"]["array"]
sr = sample["audio"]["sampling_rate"]
duration_s = len(audio_array) / sr
print(f"Audio duration: {duration_s / 60:.1f} minutes")

inputs = processor(audio=audio_array, sampling_rate=sr, return_tensors="pt", language="en")
audio_chunk_index = inputs.get("audio_chunk_index")
inputs.to(model.device, dtype=model.dtype)

start = time.time()
outputs = model.generate(**inputs, max_new_tokens=256)
text = processor.decode(outputs, skip_special_tokens=True, audio_chunk_index=audio_chunk_index, language="en")[0]
elapsed = time.time() - start
rtfx = duration_s / elapsed
print(f"Transcribed in {elapsed:.1f}s β€” RTFx: {rtfx:.1f}")
print(f"Transcription ({len(text.split())} words):")
print(text[:500] + "...")

Punctuation control

Pass punctuation=False to obtain lower-cased output without punctuation marks.

inputs_pnc = processor(audio, sampling_rate=16000, return_tensors="pt", language="en", punctuation=True)
inputs_nopnc = processor(audio, sampling_rate=16000, return_tensors="pt", language="en", punctuation=False)

By default, punctuation is enabled.

Batched inference

Multiple audio files can be processed in a single call. When the batch mixes short-form and long-form audio, the
processor handles chunking and reassembly.

from transformers import AutoProcessor, CohereAsrForConditionalGeneration
from transformers.audio_utils import load_audio

processor = AutoProcessor.from_pretrained("CohereLabs/cohere-transcribe-03-2026")
model = CohereAsrForConditionalGeneration.from_pretrained("CohereLabs/cohere-transcribe-03-2026", device_map="auto")

audio_short = load_audio(
    "https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/bcn_weather.mp3",
    sampling_rate=16000,
)
audio_long = load_audio(
    "https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/obama_first_45_secs.mp3",
    sampling_rate=16000,
)

inputs = processor([audio_short, audio_long], sampling_rate=16000, return_tensors="pt", language="en")
audio_chunk_index = inputs.get("audio_chunk_index")
inputs.to(model.device, dtype=model.dtype)

outputs = model.generate(**inputs, max_new_tokens=256)
text = processor.decode(
    outputs, skip_special_tokens=True, audio_chunk_index=audio_chunk_index, language="en"
)
print(text)

Non-English transcription

Specify the language code to transcribe in any of the 14 supported languages. This example transcribes Japanese audio from the FLEURS dataset:

from transformers import AutoProcessor, CohereAsrForConditionalGeneration
from datasets import load_dataset

processor = AutoProcessor.from_pretrained("CohereLabs/cohere-transcribe-03-2026")
model = CohereAsrForConditionalGeneration.from_pretrained("CohereLabs/cohere-transcribe-03-2026", device_map="auto")

ds = load_dataset("google/fleurs", "ja_jp", split="test", streaming=True)
ds_iter = iter(ds)
samples = [next(ds_iter) for _ in range(3)]

for sample in samples:
    audio = sample["audio"]["array"]
    sr = sample["audio"]["sampling_rate"]

    inputs = processor(audio, sampling_rate=sr, return_tensors="pt", language="ja")
    inputs.to(model.device, dtype=model.dtype)

    outputs = model.generate(**inputs, max_new_tokens=256)
    text = processor.decode(outputs, skip_special_tokens=True)
    print(f"REF: {sample['transcription']}\nHYP: {text}\n")
vLLM Integration

For production serving we recommend running via vLLM following the instructions below.

Run cohere-transcribe-03-2026 via vLLM

First install vLLM (refer to vLLM installation instructions):

uv venv --python 3.12 --seed
source .venv/bin/activate

uv pip install -U vllm==0.19.0 --torch-backend=auto
uv pip install vllm[audio]
uv pip install librosa

Start vLLM server

vllm serve CohereLabs/cohere-transcribe-03-2026 --trust-remote-code

Send request

curl -v -X POST http://localhost:8000/v1/audio/transcriptions \
 -H "Authorization: Bearer $VLLM_API_KEY" \
-F "file=@$(realpath ${AUDIO_PATH})" \
-F "model=CohereLabs/cohere-transcribe-03-2026"

Results

English ASR Leaderboard (as of 03.26.2026)

ModelAverage WERAMIEarnings 22GigaspeechLS cleanLS otherSPGISpeechTedliumVoxpopuli
Cohere Transcribe5.428.1510.849.331.252.373.082.495.87
Zoom Scribe v15.4710.039.539.611.632.811.593.225.37
IBM Granite 4.0 1B Speech5.528.448.4810.141.422.853.893.105.84
NVIDIA Canary Qwen 2.5B5.6310.1910.459.431.613.101.902.715.66
Qwen3-ASR-1.7B5.7610.5610.258.741.633.402.842.286.35
ElevenLabs Scribe v25.8311.869.439.111.542.832.682.376.80
Kyutai STT 2.6B6.4012.1710.999.811.704.322.033.356.79
OpenAI Whisper Large v37.4415.9511.2910.022.013.912.943.869.54
Voxtral Mini 4B Realtime 26027.6817.0711.8410.382.085.522.423.798.34

Link to the live leaderboard: Open ASR Leaderboard.

Human-preference results

We observe similarly strong performance in human evaluations, where trained annotators assess transcription quality across
real-world audio for accuracy, coherence and usability.
The consistency between automated metrics and human judgments suggests that the model’s improvements translate
beyond controlled benchmarks to practical transcription settings.

_Figure: Human preference evaluation of model transcripts. In a head-to-head comparison,
annotators were asked to express preferences for generations which primarily preserved meaning -
but also avoided hallucination, correctly identified named entities,
and provided verbatim transcripts with appropriate formatting.
A score of 50% or higher indicates that Cohere Transcribe was preferred on average in the comparison._

per-language WERs

Figure: per-language error rate averaged over FLEURS, Common Voice 17.0, MLS and Wenet tests sets (where relevant for a given language). CER for zh, ja, ko β€” WER otherwise

Resources

For more details and results:

Strengths and Limitations

Cohere Transcribe is a performant, dedicated ASR model intended for efficient speech transcription.

Strengths

Cohere Transcribe demonstrates best-in-class transcription accuracy in 14 languages. As a dedicated speech recognition model, it is also efficient, benefitting from a real-time factor up to three times faster than that of other, dedicated ASR models in the same size range. The model was trained from scratch, and from the outset, we deliberately focused on maximizing transcription accuracy while keeping production readiness top-of-mind.

Limitations
  • Single language. The model performs best when remaining in-distribution of a single, pre-specified language amongst the 14 in the range it supports. It does not feature explicit, automatic language detection and exhibits inconsistent performance on code-switched audio.
  • Timestamps/Speaker diarization. The model does not feature either of these.
  • Silence. Like most AED speech models, Cohere Transcribe is eager to transcribe, even non-speech sounds. The model thus benefits from prepending a noise gate or VAD (voice activity detection) model in order to prevent low-volume, floor noise from turning into hallucinations.

Ecosystem support πŸš€

Cohere Transcribe is supported on the following libraries/platforms:

If you have added support for the model somewhere not included above please raise an issue/PR!

If you find issues with any of these please raise an issue with the respective library.

Model Card Contact

For errors or additional questions about details in this model card, contact labs@cohere.com or raise an issue.

Terms of Use:
We hope that the release of this model will make community-based research efforts more accessible, by releasing the weights of a highly performant 2 billion parameter model to researchers all over the world. This model is governed by an Apache 2.0 license.

Citation

To cite this model please use the following bibtex:

@misc{julian_mack_2026,
	author       = { Julian Mack and Ekagra Ranjan and Walter Beller-Morales and Bharat Venkitesh and Pierre Richemond },
	title        = { cohere-transcribe-03-2026 (Revision d96e814) },
	year         = 2026,
	url          = { https://huggingface.co/CohereLabs/cohere-transcribe-03-2026 },
	doi          = { 10.57967/hf/8653 },
	publisher    = { Hugging Face }
}