Changelog

Inference 0.40.0 adds a Camera Calibration block for use in calibrating cameras for vision systems and speed ups to PyTorch models.

You can now train YOLOv12 models on Roboflow and deploy them using the Roboflow Serverless Hosted API.

You can now use Qwen2.5-VL with a Dedicated Deployment in Roboflow Workflows, and deploy the Workflow on your own hardware or with your Dedicated Deployment.

The Roboflow Workflows web editor now has an Ask AI button you can use as a partner in building your Workflows.

You can now train a computer vision model in minutes with Roboflow Instant.

Customers on Basic and Growth plans can now download weights from models trained with Roboflow Train. Exports are available from model pages in your Roboflow dashboard. This feature is ideal to use your model in environments that Roboflow does not yet support.

You can now use an interactive editor to draw polygon zones for use in counting and objects in a region.

Use the Llama 3.2 multimodal model in Roboflow Workflows for VQA, OCR, image classification, and more.

The latest Inference release comes with support for Llama Vision 3.2 in Roboflow Workflows, support for using MQTT in Workflows, PLC integration support with Workflows, and more.

We’ve improved the workflow editor with single-step debugging, a refreshed block design, and an integrated run preview for a smoother, more efficient build-and-test experience.