Predibase

?? Introducing the first end-to-end platform for #Reinforcement Fine-Tuning! With just a dozen labeled examples, you can fine-tune models that outperform OpenAI and #DeepSeek on complex tasks. Built on the GRPO methodology that DeepSeek-R1 popularized, our platform lets you turn any open-source LLM into a reasoning powerhouse. ?? In our real-world PyTorch to Triton transpilation case study, we achieved a 3x higher accuracy than OpenAI o1 and DeepSeek-R1 when writing GPU code – unlocking smarter, more efficient AI models. ?? What’s next? ?? Read the blog to see how it works. ?? Join us for the launch webinar on 3/27 to dive deep into RFT. ??? Try the #RFT Playground to see how it works. All links are in the comments. Let’s redefine what’s possible with fine-tuned AI! ??

AI #infra that just ain't tricky. ?? If you're at NVIDIA #GTC2025 and you want the best #models, the best #throughput and obviously, the best swag, then hit us up at booth #3011. #RunRFT #InfraSoFresh

Yesterday we launched Reinforcement Fine-Tuning! Use it to fine-tune open-source LLMs to outperform commercial models with only a handful of labeled datarows. RFT helps align models to your specific needs through reward-based learning. Curious what it looks like? Take a look at our interactive RFT playground. You can: ?? See reward scores improve over time ?? Explore each reward function ??? Inspect model generations throughout training Check it out for yourself! (?? Link in comments)

Are you at NVIDIA #GTC? Then you must stop by booth #3011 for a fit check. We have the best swag (and the best #AI infra) ?? #WalkthisWay #RunRFT

Starting today, you can build your own custom AI models without collecting labeled data – with the first end-to-end reinforcement fine-tuning platform. DeepSeek-R1 showed the world how RL can solve challenging reasoning tasks, and now we’ve baked these capabilities into an intuitive platform so anyone can leverage self-improving models for their use cases. RFT guides a model with reward functions and unlabeled data in a new interactive experience that’s a game-changer for tasks like code generation, complex RAG and more. Our early results with the platform have blown us away. We used it ourselves to build a 32B-param model to write CuDA code 3x more accurately and faster than OpenAI’s o1 or DeepSeek-R1, that we’re open sourcing today. And we’ve already seeing great traction with early customers. I’m very excited to launch the first version of this experience and give you an early look at what we’re building. Check out how it works in the thread below and let’s shape the future of custom AI together.

Who’s ready for day one of NVIDIA #GTC! If you: ? Want to train your own LLMs that beat #GPT4 ? Try out reinforcement fine-tuning with the first platform for #RFT ? Serve dozens of LLMs on a single #GPU 4x faster ?or just get some amazing swag! Then check out booth 3011 in the Inception Pavilion - we have you covered!

1000s of fine-tuned #LLMs on 1 GPU because why would you do anything else ?? LLMs as they were meant to be served with open-source #Lorax ?? Thanks for the shoutout Akshay Pachaar

?? Ready to train your own #DeepSeek-R1 with GRPO? We’ve got you covered! Predibase is the first fully managed platform for Reinforcement Fine-Tuning (#RFT)—making it easier than ever to fine-tune powerful LLMs. Don't get left behind! Join our live demo + AMA next week to see it in action. #AI #MachineLearning #LLMs #FineTuning

What does it take to build #agentic AI workflows at massive scale? ?? Check out our latest story from Marsh McLennan, one the world’s largest professional services firm, to learn how they built LenAI, an internal #copilot, that helps their 90,000 employees automate time intensive tasks. Here are a few highlights: ? Over 1 million hours saved with AI-driven #efficiency ?? Improved model #accuracy by up to 12% with fine-tuned SLMs ?? Reduced response times to mere seconds with #TurboLoRA Links in the comments ??

The future of #AI isn’t about massive, one-size-fits-all models. It’s about #specialized, domain-specific models that excel at their tasks—and reinforcement fine-tuning (#RFT) is a key to achieving peak results. Curious about advanced LLM #customization methods, how RFT works in practice, and real-world applications? Then you won't want to miss the latest episode of The Data Exchange Podcast with AI visionary Ben Lorica 罗瑞卡 and Predibase CTO & Cofounder, Travis Addair. Link in the comments ??