OpenPipe
AcquiredTurn expensive prompts into cheap fine-tuned models
About
OpenPipe is an SDK that abstracts away fine-tuning custom models. We capture your existing provider’s prompt-completion pairs in the background and use them to create a new model that is faster, cheaper and often more accurate than the original.
Founders · 2
Repeat founder and former engineer at Google and YC. Led the Startup School team and built products that increased YC applications by over 40%.
I'm a co-founder at OpenPipe, a platform for turning your slow and expensive prompts into cheap fine-tuned models. My cofounder and I wrote the first web agent that ran on GPT-4 (Taxy.AI) and I've been fine-tuning models since 2021.
Launch
OpenPipe automatically captures your existing prompts and completions, then uses them to fine-tune a drop-in replacement. The replacement is faster, cheaper and often more accurate than your original prompt.
OpenPipe captures existing prompts and completions and uses them to fine-tune a model tailored to a use-case, providing a drop-in replacement that runs faster and cheaper than the original prompts. The service enables users to fine-tune small models from hundreds to thousands of completions, then self-host or host via OpenPipe, reducing cost and latency while maintaining accuracy.
From the original launch (Aug 2023) — may be outdated.
Related startups
The API for LLM tool use
Open source tools for developing generative AI applications
