Roboflow
Day 28 after signup
New in October: Anomaly Detection, Roboflow Agent
Plus Workflow Evals, Gemini 4 Argon, and more
Why it works
- Lead with a concise overview of the most important releases before explaining each one in more detail.
- Give major features their own short description and relevant call to action so readers can explore what interests them.
- Group smaller updates into a scannable roundup and finish with a link to the complete changelog.
How it is built
- Product updates introduction
- Major feature highlights
- Feature-specific calls to action
- Additional platform updates
Email copy
Product Updates Last month we shipped Anomaly Detection to identify defects without requiring examples in advance; Roboflow Agent for designing a Workflow from a description of what you'd like to build; Workflow Evals to find failures and check for regressions before deployment; and much more. See Every October Launch Find anything that looks different from normal, like rare defects or a stray object on a manufacturing line, without collecting examples of those defects first. Train on images of good parts, then deploy to the cloud or the edge. Try Anomaly Detection Describe what you need to see, and Agent recommends the right models, designs the Workflow, and helps you get it running in the cloud or on your hardware. Meet Roboflow Agent Test Workflows against a standard dataset and compare their speed and accuracy side by side. You can also ask Agent or another AI assistant to design new Workflows, evaluate them, and report back on which one works best. Compare Your Workflows More Platform Updates Roboflow in ChatGPT → Add Roboflow from ChatGPT's plugin directory, then label, train, and deploy models without leaving the chat. New Models in Auto Label → Label object detection datasets with Gemini 4 Argon, Claude Opus 5.5, or GPT-6.1 Sol, and classification projects with GPT-6.1 Sol. New Models in Workflows → Run Mistral Large 4 with the new Mistral AI block, or add Claude Haiku 5.5, Muse Wonder, and Gemini 4 Argon as steps for detection, OCR, and visual question answering. Mirror Videos from Your Cloud Buckets → Datasources now pull video from AWS S3, Google Cloud Storage, and Azure Blob Storage. RF-DETR Semantic Segmentation → Train segmentation models in six sizes, from Nano to 2X Large, and let Neural Architecture Search find the best fit for your data. Enforce Your Own LLM Keys → Run Agent, Auto Label, and Workflow blocks on your OpenAI, Azure OpenAI, or Gemini keys. View the Full Changelog © 2026 Roboflow, Inc. All rights reserved. 500 Locust St, Suite 104, Des Moines, IA 50309 Manage Preferences | Unsubscribe