Episode Details

Hey all, Alex here 👋

This week, while not the busiest week in releases (we can't get a SOTA LLM every week now can we), was full of interesting open source releases, and feature updates such as the chatGPT meetings recorder (which we live tested on the show, the limit is 2 hours!)

It was also a day after our annual W&B conference called FullyConnected, and so I had a few goodies to share with you, like answering the main question, when will W&B have some use of those GPUs from CoreWeave, the answer is... now! (We launched a brand new preview of an inference service with open source models)

And finally, we had a great chat with Pankaj Gupta, co-founder and CEO of Yupp, a new service that lets users chat with the top AIs for free, while turning their votes into leaderboards for everyone else to understand which Gen AI model is best for which task/topic. It was a great conversation, and he even shared an invite code with all of us (I'll attach to the TL;DR and show notes, let's dive in!)

00:00 Introduction and Welcome

01:04 Show Overview and Audience Interaction

01:49 Special Guest Announcement and Experiment

03:05 Wolfram's Background and Upcoming Hosting

04:42 TLDR: This Week's Highlights

15:38 Open Source AI Releases

32:34 Big Companies and APIs

32:45 Google's Gemini Updates

42:25 OpenAI's Latest Features

54:30 Exciting Updates from Weights & Biases

56:42 Introduction to Weights & Biases Inference Service

57:41 Exploring the New Inference Playground

58:44 User Questions and Model Recommendations

59:44 Deep Dive into Model Evaluations

01:05:55 Announcing Online Evaluations via Weave

01:09:05 Introducing Pankaj Gupta from YUP.AI

01:10:23 YUP.AI: A New Platform for Model Evaluations

01:13:05 Discussion on Crowdsourced Evaluations

01:27:11 New Developments in Video Models

01:36:23 OpenAI's New Transcription Service

01:39:48 Show Wrap-Up and Future Plans

Here's the TL;DR and show notes links

ThursdAI - June 19th, 2025 - TL;DR

* Hosts and Guests

* Alex Volkov - AI Evangelist & Weights & Biases (@altryne)

* Co Hosts - @WolframRvnwlf @yampeleg @nisten @ldjconfirmed

* Guest - @pankaj - co-founder of Yupp.ai

* Open Source LLMs

* Moonshot AI open-sourced Kimi-Dev-72B (Github, HF)

* MiniMax-M1 456B (45B Active) - reasoning model (Paper, HF, Try It, Github)

* Big CO LLMs + APIs

* Google drops Gemini 2.5 Pro/Flash GA, 2.5 Flash-Lite in Preview ( Blog, Tech report, Tweet)

* Google launches Search Live: Talk, listen and explore in real time with AI Mode (

Listen Now