The Cerebras Wafer Scale Engine

OpenAI’s o1 Model: A Breakthrough in AI Reasoning

OpenAI launched its 01 model, also known as Strawberry, on September 12, featuring a full-function o1-preview and a more affordable o1-mini to demonstrate how AI can be greatly improved by breaking a query down into step-by-step reasoning. This chain-of-thought reasoning works more like humans do, decomposing a problem into bite-sized chunks and tackling those chunks in sequence. Especially useful in STEM and math problems, o1 generated a lot of interest.

Is o1 That Good?

OpenAI’s o1 is certainly impressive. While most would argue it isn’t yet general AI, it scores 83% on the Mathematics Olympiad, basically operating at a Ph.D.-level of accuracy, vs. 13% for GPT4.0.

However, o1 has three problems: it is very slow; it is four times more expensive ($60 per million tokens vs. $15 for GPT4); and it is text only. Solving the last problem won’t be that difficult for OpenAI. And it can make progress on the first two, but the step-wise problem solving and subsequent iteration to find the “best” answer just takes a ton of compute. That isn’t easy to solve.

Cerebras Announces o1-Like Reasoning On Llama 3.3-70B

Now the first two issues are being worked on by Cerebras Systems. Cerebras announced today at the annual NeurIps conference that it has open-sourced a version of Llama called CePO (Cerebras Planning and Optimization) based on Llama 3.3-70B that adds o1-like reasoning to Llama models. Cerebras claims that this is the first and only reasoning model that is able to run in real-time, providing answers to complex questions in seconds instead of the minutes required by OpenAI’s o1 model.

CePO enables Meta’s Llama 70B to be more accurate than Llama 405B and for many problems more accurate than GPT4. Cerebras has not yet productized this new model, but may add it to its inference service portfolio announced earlier this year.

What’s Our Take?

The 01-style of reasoning is quite powerful, but requires far more computation than GPT 4.0. Based on the current OpenAI pricing model, it probably requires something like four times more compute power. That’s where Cerebras’ Wafer-Scale Engine comes in, with tons of cores and SRAM memory on the frisbee-sized “chip.”

Conclusion

Cerebras’ Wafer-Scale Engine and CePO model are promising developments in the field of AI reasoning. With the ability to run in real-time and provide accurate answers to complex questions, CePO has the potential to revolutionize the way we approach AI-powered decision-making.

Frequently Asked Questions

Q: What is the Cerebras Wafer-Scale Engine?
A: The Cerebras Wafer-Scale Engine is a processing chip designed to handle complex AI computations.

Q: What is CePO?
A: CePO (Cerebras Planning and Optimization) is an open-sourced version of Llama that adds o1-like reasoning to Llama models.

Q: What is the advantage of CePO over other AI models?
A: CePO is the first and only reasoning model that is able to run in real-time, providing answers to complex questions in seconds instead of the minutes required by other models.

Q: How does CePO compare to OpenAI’s o1 model?
A: CePO is more accurate and faster than OpenAI’s o1 model, with the ability to provide accurate answers to complex questions in real-time.