Cohere has released Transcribe Arabic, an open-source model for Arabic speech recognition available on Hugging Face. The 2-billion-parameter model outperforms competitors like Whisper and OmniASR on dialect recognition, code-switching, and bilingual Arabic-English speech.
Transcribe Arabic addresses gaps in existing speech-to-text systems that struggle with Arabic's linguistic complexity. The model handles regional dialects, code-switching between languages, and mixed Arabic-English speech—challenges that have limited performance in previous solutions.
The model is distributed under the Apache 2.0 license, enabling researchers and developers to use and modify it freely. Its availability on Hugging Face provides easy access for integration into applications and further development.
Arabic speech recognition has lagged behind English-language models, partly due to the language's phonetic complexity and the fragmentation of dialects across regions. Cohere's release addresses this gap by providing a specialized tool built specifically for Arabic's unique transcription requirements.
The 2-billion-parameter model represents a middle ground in scale—larger than many specialized models but smaller than massive general-purpose systems, balancing performance with practical deployment considerations.
Researchers have developed a typeface that remains readable to humans while evading optical character recognition systems. The font exploits weaknesses in how machine learning models process visual text.
Cosmetic surgeons report a growing number of patients requesting surgery based on AI-generated images, often with unrealistic expectations about achievable results.
US software development job postings on Indeed jumped 15% since Claude Code launched in February 2025, defying a broader 7% decline in overall job listings, according to Indeed Hiring Lab data.
Beijing Academy of Artificial Intelligence released Orca, a world model trained on 125,000 hours of unlabeled video that matches specialized robotics systems without ever seeing a single action label.