So-called reasoned AI models are increasingly easier – and less expensive – to develop.
On Friday, NovaSky, a team of researchers based at UC Berkeley’s Sky Computing Lab, released Sky-T1-32B-Preview, a competitive reasoning model with an earlier version of OpenAI’s o1 on a number of tests keys. Sky-T1 appears to be the first truly open source reasoning model in the sense that it can be replicated from scratch; the team released the dataset they used to train it along with the necessary training code.
“Remarkably, Sky-T1-32B-Preview was formed for less than $450,” the team wrote in a statement. blog post“demonstrating that it is possible to reproduce high-level reasoning skills in an affordable and effective manner.”
Unlike most AI, reasoning models do their own fact-checking, which helps them avoid some of the pitfalls that normally cause models to fail. Reasoning models take a little longer (usually seconds to minutes) to arrive at solutions compared to a traditional model without reasoning. The upside is that they tend to be more reliable in areas like physics, science, and math.
The NovaSky team claims to have used another reasoning model, Alibaba’s QwQ-32B-Preview, to generate the initial training data for Sky-T1, then “curate” the data mix and leverage GPT-4o -mini of OpenAI to refactor data into a more workable format. Training the Sky-T1, with 32 billion parameters, took approximately 19 hours using a rack of 8 Nvidia H100 GPUs. (The settings roughly correspond to a model’s problem-solving skills.)
According to the NovaSky team, Sky-T1 performs better than an early version of o1 on MATH500, a collection of “competition-level” math challenges. The model also beats o1’s preview on a hard problem set from LiveCodeBench, a coding assessment.
However, Sky-T1 does not measure up to the o1 overview on GPQA-Diamond, which contains questions related to physics, biology and chemistry that a PhD holder would be expected to know.
It’s also important to note that o1’s GA version of OpenAI is a more powerful model than o1’s pre-release version, and OpenAI is expected to release an even more capable reasoning model, o3, in the coming weeks .
But the NovaSky team says Sky-T1 marks just the beginning of its journey to develop open source models with advanced reasoning capabilities.
“In the future, we will focus on developing more efficient models that maintain strong reasoning performance and exploring advanced techniques that further improve model efficiency and accuracy at test time,” wrote the team in the message. “Stay tuned as we move forward on these exciting initiatives. »
