Weights & Biases

Introducing W&B Serverless SFT: Fine-tune LLMs with no infra headaches

Product newsletter: Updates and new features for June 2026

AI agent observability

• Evaluations: Evaluations
• Agents: Agents
• Inference: Inference
• Safety and scoring: Safety and scoring

Build models

• Computer Vision: Computer Vision
• NLP: NLP
• MLOps: MLOps
• LLM: LLM

About Weights & Biases

• Integrations: Integrations
• W&B releases: W&B releases
• Events: Events
• Podcast: Podcast

Community Posts

• Track, guard, and remediate AI usage with W&B Weave and CrowdStrike Falcon AIDR
  Learn how Weave and CrowdStrike Falcon AI Detection and Response can help ensure internal AI systems are handling sensitive data appropriately
  Read more

• Tabular ML didn't die. It was just waiting for agents
  Learn how our tabular data agent took #1 on the MLE-bench tabular slice with two golds, three above-median, four-for-four valid submissions.
  Read more

• Introducing CoreWeave ARIA: AI Research and Iteration Agent
  Your experiments are already tracked. Now let the agent read them, analyze them, and help you turn every experiment into continuous improvement.
  Read more

• We'll be at the AI Engineer World's Fair - Come Say Hi
  Join Weights & Biases by CoreWeave at AI Engineer World’s Fair in SF. Visit booth UG 24 for demos, swag, speaker sessions, and new Weave agent tracing features.
  Read more

• How to structure effective Agent Skills and evaluate whether they actually work
  Improve your Vibe Coding Skills using Agent Skills and evaluate them with Skills Bench and W&B Weave
  Read more

• ART-Optimized Megatron (AOM): 12x higher training throughput
  How we massively improved training throughput on our open-source ART training library
  Read more

• Building a production-grade MLOps pipeline for VLA models
  This report covers the development of a compact vision-language-action model built around Gemma 3 270M.
  Read more

• Getting started with embodied chain-of-thought (ECoT)
  Learn how to get started with Embodied Chain-of-Thought, a robotics reasoning approach that helps vision-language-action models plan, explain, and execute grounded actions.
  Read more

• Governance workflows for AI agents: an evidence-backed review gate on W&B Weave
  A technical walkthrough of an open-source AI governance toolkit that turns scattered evals, red-team runs, and approvals into one reproducible review gate, with every finding traced in W&B Weave.
  Read more

• Tutorial: Benchmarking Claude Opus 4.8 honesty with BeHonest and W&B Weave
  How honest is Claude Opus 4.8 compared to 4.7? We ran both models through the BeHonest benchmark and logged every result into W&B Weave.
  Read more

• New in W&B Weave: Observability and continuous improvement for production agents
  We're shipping a new set of capabilities that make Weave the best tool to raise and maintain agent quality in production. Here's what you need to know.
  Read more

• Build a Robotics Data Flywheel on CoreWeave using NVIDIA Cosmos 3
  A foundation is large-scale synthetic data and compute, a requirement that is no longer an issue when teaming up with CoreWeave and NVIDIA.
  Read more

• Product newsletter: Updates and new features for May 2026
  From a brand new CoreWeave Sandboxes offering to updates in Models and our iOS app, here's what we released in May.
  Read more

• Running agents in production with Google’s Gemini Managed Agents
  In this article, we will use the Gemini Managed Agents API to build a small bug-finder agent.
  Read more

• Building a text-to-SQL agent
  In the second edition of our AI Builders series, learn how to build a text-to-SQL agent
  Read more

• Introducing the W&B MCP Server: An agent-native interface for your experiments and traces
  Our MCP server is generally available now. Built as primitives so the agent can do the rest.
  Read more

• Product newsletter: Updates and new features for April 2026
  From universally available automations to brand new Weave dashboards, here are the big features we released in April
  Read more

• Training a Gemma-3 powered VLA model
  This report covers the development of a compact vision-language-action model built around Gemma 3 270M.
  Read more

• From sequence to structure: Tracking protein fold prediction with W&B
  Learn how to track protein fold prediction using ESMFold and Weights & Biases. Explore a complete biotech pipeline, from 3D structure visualization and pLDDT tracking to Bayesian sweeps and model versioning.
  Read more

• Building an AI agent for interior design
  In the inaugural edition of our AI Builders series, learn how to build an AI agent for home decor
  Read more

• Tutorial: Building a production-ready fraud triage Copilot
  Build, evaluate, guard, and monitor an LLM fraud analyst end-to-end with W&B Weave.
  Read more