W&B Weave newsletter: General availability, playground, guardrails, leaderboards, and more
We made a lot of improvements to Weave in the last month. Here's what you need to know
Created on December 6|Last edited on December 6
Comment
We have some exciting news to share: W&B Weave is now generally available on SaaS and AWS dedicated cloud. This comes alongside a slew of new Weave features available as of December 2nd. In the video below, our CEO Lukas Biewald shows you how to build agentic AI applications using some of our latest Weave features, including leaderboards and annotation templates.
But before diving into all the news features, let's kick things off with our LLM tip of the week.
LLM tip of the week ✅
If you haven't explored the reasoning models when o1-preview came out, now may be a good time to dive back in again. Recent releases such as Deepseek R1 and the open sourced QwQ, confirmed the existence of a new kind of scaling law, called test-time compute, showing that the longer models have time to "think," the better they are at answering questions. Effectively, this allows a 32B parameter model to outperform a 405B LLama on many reasoning and math tasks while running on consumer hardware like a Macbook Pro.
W&B Weave news 🚀
Leaderboards
Weave now allows developers to group evaluations into leaderboards and share them across their organization.

Playground
To evaluate models and prompts without jumping into code, Weave offers a playground to quickly iterate on prompts and see how the LLM response changes.

Guardrails (currently in preview)
Due to the non-deterministic nature of LLMs, AI can sometimes behave inappropriately or leak private data. Guardrails offers out-of-the-box filters to detect harmful outputs and prompt attacks. Once an issue is detected, pre and post hooks help trigger safeguards.
Evaluations (currently in preview)
With online evaluations, you can score live traces from production to monitor your application in real time. Online evaluations allow developers to separate evaluations from core application processing.
Community 🏡
W&B Weave now has a dedicated home on 𝕏 at @weave_wb. We’ll keep you up to date on Weave product developments, share LLM tips and tricks, show best practices for production GenAI applications, and more. Give us a follow.
Also, with the release of Weave GA, Alex Volkov (our AI Evangelist) shared a simplified four step cookbook for productizing LLM applications with confidence on 𝕏. Head ****here to check it out.
Popular articles 📑
In honor of AWS re:Invent 2024, we published an article about how to use Weights & Biases with Amazon Bedrock:
Compare LLMs on Amazon Bedrock for text summarization with W&B Weave
Discover how to use Amazon Bedrock in combination with Weave to evaluate and compare LLMs for summarization tasks, leveraging Bedrock’s managed infrastructure and Weave’s advanced evaluation features.
Events 🗓️
Join our very own Director of AI Morgan McGuire as he presents a workshop on how to train, fine-tune, and manage models from experimentation to production.
During NeurIPS we will be hosting a happy hour with Lambda and Gradient Ventures. We’d love it if you joined us.
This event will be held at our headquarters where we’ll explore best practices for creating adaptive AI-generated UIs. We’ll touch on insights from industry experts on dynamic interface design, engineering challenges, and implementation strategies.
GenAI master class: To Production Series. Part 3: Build and deploy a GenAI app end-to-end- Dec 16 at 5:30pm in SF
Join us in person in our San Francisco office for part 3 of our master class all about equipping you with the practical skills to build and deploy GenAI solutions.
Need help getting started with W&B Weave?
Add a comment
Iterate on AI agents and models faster. Try Weights & Biases today.