Cohere's open-weight ASR model hits 5.4% word error rate — low enough to replace speech APIs in production pipelines

Source: VentureBeat

Enterprises building voice‑enabled workflows have had limited options for production‑grade transcription: closed APIs with data residency risks, or open models that trade accuracy for deployability. Cohere’s new open‑weight ASR model, Transcribe, is built to compete on all four key differentiators — contextual accuracy, latency, control, and cost.

Cohere says that Transcribe outperforms current leaders on accuracy — and, unlike closed APIs, it can run on an organization’s own infrastructure. The model is available via an API or in Cohere’s Model Vault as cohere-transcribe-03-2026, has 2 billion parameters, and is licensed under Apache‑2.0. Cohere reports an average word error rate (WER) of 5.42 %, fewer mistakes than similar models.

It’s trained on 14 languages: English, French, German, Italian, Spanish, Greek, Dutch, Polish, Portuguese, Chinese, Japanese, Korean, Vietnamese, and Arabic (the specific Chinese dialect was not disclosed). Cohere emphasized a “deliberate focus on minimizing WER, while keeping production readiness top‑of‑mind.” The result is a model that enterprises can plug directly into voice‑powered automations, transcription pipelines, and audio‑search workflows.

Self‑hosted transcription for production pipelines

Until recently, enterprise transcription has been a trade‑off — closed APIs offered accuracy but locked in data; open models offered control but lagged on performance. Unlike Whisper, which launched as a research model under an MIT license, Transcribe is available for commercial use from release and can run on an organization’s own local GPU infrastructure. Early users flagged the commercial‑ready open‑weight approach as meaningful for enterprise deployments.

Organizations can bring Transcribe to their own local instances, as Cohere notes the model has a more manageable inference footprint for local GPUs. The company attributes this to extending “the Pareto frontier, delivering state‑of‑the‑art accuracy (low WER) while sustaining best‑in‑class throughput (high RTFx) within the 1B+ parameter model cohort.”

How Transcribe stacks up

Transcribe outperformed speech‑model stalwarts, including Whisper from OpenAI (which powers the voice feature of ChatGPT) and ElevenLabs (widely used by large retail brands). It currently tops the Hugging Face ASR leaderboard with an average WER of 5.42 %, beating Whisper Large v3 (7.44 %), ElevenLabs Scribe v2 (5.83 %), and Qwen3‑ASR‑1.7B (5.76 %).

Based on other datasets tested by Hugging Face, Transcribe also performed well:

• AMI dataset (meeting understanding and dialogue analysis): 8.15 %
• Voxpopuli dataset (accent comprehension): 5.87 %, beaten only by Zoom Scribe.

Early users have highlighted accuracy and local deployment as standout factors — especially for teams that have been routing audio data through external APIs and now want to bring that workload in‑house.

For engineering teams building RAG pipelines or agent workflows with audio inputs, Transcribe offers a path to production‑grade transcription without the data residency and latency penalties of closed APIs.