Sourced from mlflow's releases.

3️⃣ MLflow 3 3️⃣

MLflow 3: Redefining MLOps for the GenAI Era

MLflow 3 is now available to everyone, marking the biggest evolution in the best open-source MLOps platform's history and transforming how millions of developers build, deploy, AI applications. While previous versions focused on traditional ML workflows, MLflow 3 fundamentally reimagines the platform for the GenAI era. This isn't just an update, but a complete paradigm shift that brings enterprise-grade GenAI capabilities to the open source community for the first time.

🎯 Improved Model Tracking for GenAI

MLflow 3 introduces a refined architecture with the new LoggedModel entity as a first-class citizen, moving beyond the traditional run-centric approach. This enables better organization and comparison of GenAI models. agents, deep learning checkpoints, and model variants across experiments.

🔗 Comprehensive Performance Tracking & Observability

Enhanced model tracking provides comprehensive lineage between models, runs, traces, prompts, and evaluation metrics. The new model-centric design allows you to group traces and metrics from different development environments and production, enabling rich comparisons across model versions.

📊 Production-Grade GenAI Evaluation

MLflow's evaluation and monitoring capabilities help you systematically measure, improve, and maintain the quality of your GenAI applications throughout their lifecycle. From development through production, use the same quality scorers to ensure your applications deliver accurate, reliable responses while managing cost and latency. Visit documentation for more details.

👥 Human-in-the-Loop Feedback

Real-world GenAI applications need human oversight. MLflow 3 now tracks human annotations and feedback for model predictions, enabling streamlined human-in-the-loop evaluation cycles. This creates a collaborative environment where data scientists, domain experts, and stakeholders can efficiently improve model quality together. (Note: Currently available in Databricks Managed MLflow. Open source release coming in the next few months.)

⚡️ State-of-the-Art Prompt Optimization

Transform prompt engineering from art to science. The MLflow Prompt Registry now includes prompt optimization capabilities built on top of the state-of-the-art research, allowing you to automatically improve prompts using evaluation feedback and labeled datasets. This includes versioning, tracking, and systematic prompt engineering workflows.

📚 Revamped Website and Documentation

The MLflow documentation and website has been fully redesigned to support two main user journeys: GenAI development and classic machine learning workflows. The new structure offers dedicated sections for GenAI features (including LLMs, prompt engineering, and tracing), and traditional ML capabilities such as experiment tracking, model registry, deployment, and evaluation.

▶︎▶︎▶︎ Ready to Get Started?　▶︎▶︎▶︎

Get up and running with MLflow 3 in minutes:

pip install 'mlflow>=3.1'

Resources:

🌐 New Website | 📖 Documentation | 🎉:Release Notes

🏎️ The Road Ahead 🏎️

It is just the beginning. The open source community continues driving innovation toward the world's best open-source MLOps/LLMOps platform. Here's how you can be part of the journey:

How to Get Involved:

• 🔧 Contribute Code: From bug fixes to major features, all contributions welcome
• 🐝 Report Issues: Help us improve by reporting bugs and requesting features
• 💬 Join Discussions: Technical discussions, roadmap planning, and peer support
• 📝 Share Your Story: Write blogs, tutorials, and docs about your MLflow implementations to help others learn!

... (truncated)

Sourced from mlflow's changelog.

CHANGELOG

3.1 (2025-06-11)

MLflow 3 includes several major features and improvements

Features:

• [Tracking] MLflow 3.0 (#13211, @​harupy)
• [Prompts] Add Custom Prompt Judges to mlflow[databricks] (#16097, @​dbrx-euirim)
• [Artifacts / Model Registry / Tracking] Package model environment when registering model (#15783, @​qyc)
• [Tracking] Add MlflowSparkStudy (#15418, @​lu-wang-dl)
• [Scoring] Make spark_udf support DBConnect + DBR 15.4 / DBR dedicated cluster (#15968, @​WeichenXu123)
• [Tracking] Lock model dependencies when logging a model using uv (#15875, @​harupy)
• [Model Registry] Introduce mlflow.genai.optimize_prompt to optimize prompts (#15861, @​TomeHirata)
• [Tracing] Support custom request/response preview (#15919, @​B-Step62)
• [Tracking] Add integration for AutoGen > 0.4 (#14729, @​TomeHirata)
• [Tracking] Support token tracking for OpenAI (#15870, @​B-Step62)
• [Tracking] Support tracing ResponsesAgent.predict_stream (#15762, @​bbqiu)
• [Tracking] Introduce client and fluent APIs for LogLoggedModelParams (#15717, @​artjen)
• [Models] Support predict_stream in DSPy flavor (#15678, @​TomeHirata)
• [Tracking] Record notebook and git metadata in trace metadata (#15650, @​B-Step62)
• [Model Registry] Added search_prompts function to list all the prompts registered (#15445, @​joelrobin18)
• [Models] Support compression for pyfunc log model (#14700, @​antbbn)
• [Gateway] Add support for Gemini in AI Gateway (#15069, @​joelrobin18)
• [Tracing] PydanticAI Autologging (#15553, @​joelrobin18)
• [Tracking] Support setting databricks auth profile by DATABRICKS_CONFIG_PROFILE environment variable. (#15587, @​WeichenXu123)
• [Tracking] create mlflow tracing for smolagents (#15574, @​y-okt)
• [Artifacts / UI] Support for video artifacts (#15518, @​joelrobin18)
• [Model Registry] Add allow_missing parameter in load_prompt (#15371, @​joelrobin18)
• [Tracking] Emit a warning for mlflow.get_artifact_uri() usage outside active run (#12902, @​Shashank1202)

Bug fixes:

• [GenAI] Add Databricks App resource (#15867, @​aravind-segu)
• [Tracking] Support json-string for inputs/expectations column in Spark Dataframe (#16011, @​B-Step62)
• [Tracking] Avoid generating traces from scorers during evaluation (#16004, @​B-Step62)
• [GenAI] Allow multi inputs module in DSPy (#15859, @​TomeHirata)
• [Tracking] Improve error handling if tracking URI is not set when running mlflow gc (#11773, @​oleg-z)
• [Tracking] Trace search: Avoid spawning threads for span fetching if include_spans=False (#15634, @​dbczumar)
• [Tracking] Fix global_guideline_adherence (#15572, @​artjen)
• [Model Registry] Log Resources from SystemAuthPolicy in CreateModelVersion (#15485, @​aravind-segu)
• [Models] ResponsesAgent interface update (#15601, #15741, @​bbqiu)

Breaking changes:

• [Tracking] Move prompt registry APIs under mlflow.genai.prompts namespace (#16174, @​B-Step62)
• [Model Registry] Default URI to databricks-uc when tracking URI is databricks & registry URI is unspecified (#16135, @​dbczumar)
• [Tracking] Do not log SHAP explainer in mlflow.evaluate (#15827, @​harupy)
• [Tracking] Update DataFrame schema returned from mlflow.search_trace() to be V3 format (#15643, @​B-Step62)

... (truncated)

• 39a419b Run python3 dev/update_mlflow_versions.py pre-release ... (#16187)
• c44e13f Run python3 dev/update_ml_package_versions.py (#16186)
• 43b4091 Mlflow 3 docs refactor (#15954)
• 4b74195 Remove log_models from openai autolog (#16178)
• 3d6758f Fix labeling schemas usage (#16177)
• 23595f4 [BUG] ERROR mlflow.server: Exception on /graphql when trying to open a run if...
• 6464d1b Link prompts to traces when loaded via fluent API (#16167)
• 5795d77 Update ML package versions for 3.1.0 (#16171)
• a61080e Correct the way to check the error messages for optuna study (#16169)
• e1dfdeb Move prompt registry APIs under mlflow.genai.prompt namespace (#16174)
• Additional commits viewable in compare view

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot merge will merge this PR after your CI passes on it
• @dependabot squash and merge will squash and merge this PR after your CI passes on it
• @dependabot cancel merge will cancel a previously requested merge and block automerging
• @dependabot reopen will reopen this PR if it is closed
• @dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)
  You can disable automated security fix PRs for this repo from the Security Alerts page.