Cohere has announced the retirement of fine-tuning capabilities for several models, effective April 4th, 2026. This includes models such as embed-english-v2.0, embed-english-light-v2.0, embed-multilingual-v2.0, and c4ai-aya-expanse-8b. The company states that this move is part of its ongoing effort to introduce safer and more capable models, which may require applications relying on Cohere’s models to undergo updates to maintain functionality.

Cohere has a history of deprecating older models and endpoints as newer versions become available. This process is intended to ensure that users have access to the most advanced and secure tools. Impacted customers are notified via email, documentation, and blog posts, with detailed information provided on the deprecation timeline and recommended replacements for each affected model.

The retirement of these models will take effect on April 4th, 2026, and users are given a 90-day window from the announcement to make necessary adjustments. This period allows developers and businesses to transition smoothly to the newer models, minimizing disruption to their operations. Cohere emphasizes that all affected users will be informed well in advance to facilitate the update process.

The retirement of these models may lead to increased costs for users who need to migrate their applications to newer versions. It could also result in potential vendor lock-in if users are unable to find suitable alternatives. Additionally, the transition may raise concerns about governance and the long-term sustainability of relying on Cohere’s evolving model ecosystem.

Despite the deprecation, Cohere continues to develop new models and capabilities, ensuring that users have access to the latest advancements in AI. The company remains committed to providing clear guidance and support throughout the transition period. Users are encouraged to refer to Cohere’s documentation and deprecation history for detailed information on the changes and their implications.