Cohere

A small 30B-A3B coding model by Cohere released under Apache 2.0.

Cohere, which is becoming more of a regular entrant into Artifacts lately, released their flagship, Command A+, under Apache 2.0. Previous iterations of the series have been released under a non-commercial license, so this change is more than welcome! Command A+ combines multi-modal, multi-lingual and agentic capabilities as a 218B-A25B MoE, making it usable with a single B200 (when using 4-bit).

A speech-to-text model by Cohere based on the conformer architecture, similar to NVIDIA's Parakeet. It features 14 different languages, including some AIPAC languages and Arabic. Performance-wise, Cohere claims it beats similarly sized open and closed models. To top it all off: The model is released under Apache 2.0! Previous open models by Cohere were released under a non-commercial license.

A non-commercial, small, multilingual model.

A model specialized for translation tasks. Similar to Command A reasoning, this model is released under a non-commercial license.

As usual, Cohere releases their flagship models on HuggingFace under a non-commercial license for academics to use.

The new series by Cohere, replacing the previous Command R series of models. Like other releases, Cohere releases their models under the CC-by-NC license for research purposes.

The first VLM release by Cohere, focusing on multilinguality. The model combines Cohere's Command R with SigLIP 2. Here's their win rate plot comparing peer models on multilingual, multimodal queries.

An 8B model trained for the Arabic and English languages. Similar to other Cohere releases, it is released under the CC-by-NC license.

A new model in the Command R series, which marks also the end of this model series.

Cohere has released a new model of their Aya series by fine-tuning their Command R models, focusing on multilinguality by using multilingual preference training.

Cohere updated their original Aya model with fewer languages and using their own base model (Command R, while the original model was trained on top of T5).