Cohere released Parse (parse-v5.0), a 2.3B-parameter multimodal model that converts complex documents into clean, structured Markdown for downstream AI workflows. It extracts text in reading order, tables, lists, forms, images and captions, page boundaries, and visual element locations. Outputs include Markdown/HTML content, HTML-formatted tables, bounding boxes, and image descriptions.

Cohere released Cohere Transcribe Arabic, an open-source speech-to-text model fine-tuned from Cohere Transcribe using Arabic speech data. It supports all major Arabic dialects and English spoken with an Arabic accent, and is described as the most accurate open-source Arabic ASR model available. It is optimized for production inference and throughput with a 2B parameter conformer-based encoder-decoder architecture.

Cohere released North Mini Code, their first agentic coding model. It is a 30 billion total / 3 billion active parameter Mixture of Experts model trained specifically for agentic coding tasks. The model is small enough to run on local hardware and is available via the Chat V2 API and as open weights on Hugging Face under Apache 2.0 license.