Weights & Biases Lands $50M Investment at $1.25B Valuation After Reaching 700K Users
Weights & Biases Lands $50M Investment at $1.25B Valuation After Reaching 700K Users
The generative AI tooling market valuations are following foundation models' rise
Weights and Biases (W&B), the MLOps platform for managing AI models, today announced a $50 million funding round at a new valuation of $1.25 billion. The round was led by Daniel Gross, the founder of Cue, which Apple acquired, and Nat Friedman, the former CEO of GitHub. Sapphire Ventures also participated as did investors from earlier rounds.
The company said its user base grew from 100,000 at its last funding round in 2021 to 700,000 today. Aleph Alpha, Cohere, Meta, MosaicML, NVIDIA, OpenAI, Toyota, and Volkswagen are all listed as customers. W&B also introduced a new product called Prompts for LLMs.
Gross and Friedman commented:
“Weights & Biases found their footing by helping developers supervise model training, where they are the popular choice for practically everyone we know -- from Meta to OpenAI to Microsoft. Now that the industry is maturing, they're expanding from training to prompting – helping enterprises around the world use LLM platforms in a safe & reliable way.”
Prompts for LLMs
The new W&B Prompts solution extends the company’s MLOps “experiments” capability to include an LLM debugging tool that enables users to evaluate errors and chains more easily. It also can display what W&B calls a chain topology, which describes “the settings, tools, agents, and prompt details” of a chain in an LLM operation.
Finally, and maybe most interestingly, the new Prompts product runs evaluations from OpenAI Evals - the open-source repository of LLM evaluation templates. Prompts “logs the evaluation in W&B Tables for easy visualization and analysis, and generates a Report for seamless collaboration.”
Model evaluation will be an increasingly important task as production LLM-based applications proliferate. I assume this will also support custom OpenAI Eval templates so specialized evaluations can be executed that are use-case specific. According to the company:
W&B Prompts allows LLM builders to better understand every step of their LLM programs so they can fine-tune, prompt engineer, and debug the latest foundational models. This is a critical step, enabling faster time to production on LLM models and the apps they power.
The Rise of Generative AI Tooling
If you follow generative AI, you have likely heard a lot about LLMs and other foundation models. You also have no doubt heard about generative AI-enabled applications, and maybe some GPU products and cloud services for training models. You probably have not heard much about tooling providers. That is about to change.
Scale AI, Snorkel, Langchain, W&B, and other tooling providers are critical to advancing generative AI technology and applications. These represent the picks and shovels that will ultimately transform the raw technological capabilities of generative AI foundation models into practical and reliable solutions.
