Fireworks AI Deprecates Kimi K2.5 and Qwen 3.6 Plus Serverless Models, Recommends Migrations

Fireworks AI, a prominent provider in the AI infrastructure space, has issued an important update regarding its serverless model offerings. As of June 26, 2026, the Kimi K2.5 and Qwen 3.6 Plus models are officially deprecated from their serverless platform. This means that users currently relying on these specific model versions for their applications will need to take action to avoid service disruption. The deprecation is part of the natural progression in AI model development, where newer, more advanced, and often more efficient versions are released. Fireworks AI is recommending that all users migrate their workloads to the successor models: Kimi K2.6 and Qwen 3.7 Plus. These newer iterations are expected to offer improved performance, enhanced features, or better stability, aligning with the company's commitment to providing cutting-edge AI capabilities. For organizations leveraging these models, this announcement underscores a critical aspect of MLOps: continuous model lifecycle management. In a rapidly evolving field like AI, models are not static assets. They require regular updates, performance monitoring, and, eventually, planned deprecation and migration to newer versions. This process ensures that AI systems remain robust, scalable, and capable of delivering optimal results. MLOps teams are now tasked with assessing their current deployments, planning the migration process, and validating the performance of their applications with the Kimi K2.6 and Qwen 3.7 Plus models. This typically involves updating dependencies, re-testing integrations, and potentially retraining or fine-tuning models if significant architectural changes have occurred between versions. The goal is to facilitate a smooth transition with minimal impact on production environments. The update also serves as a reminder for developers and MLOps engineers to build resilient systems that can accommodate model changes. Strategies such as version control for models, automated testing pipelines, and robust deployment mechanisms are crucial for managing such transitions effectively. By proactively addressing these updates, users can continue to benefit from Fireworks AI's innovations while maintaining the operational integrity of their AI-powered solutions.