Past benchmarks: How DeepSeek-R1 and o1 carry out on real-world duties

DeepSeek-R1 has absolutely created a variety of pleasure and concern, particularly for OpenAI’s rival mannequin o1. So, we put them to check in a side-by-side comparability on a couple of easy knowledge evaluation and market analysis duties. 

To place the fashions on equal footing, we used Perplexity Professional Search, which now helps each o1 and R1. Our objective was to look past benchmarks and see if the fashions can really carry out advert hoc duties that require gathering data from the net, selecting out the best items of knowledge and performing easy duties that may require substantial handbook effort. 

Each fashions are spectacular however make errors when the prompts lack specificity. o1 is barely higher at reasoning duties however R1’s transparency provides it an edge in instances (and there might be fairly a couple of) the place it makes errors.

Here’s a breakdown of some of our experiments and the hyperlinks to the Perplexity pages the place you possibly can overview the outcomes your self.

Calculating returns on investments from the net

Our first check gauged whether or not fashions may calculate returns on funding (ROI). We thought of a situation the place the consumer has invested $140 within the Magnificent Seven (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla) on the primary day of each month from January to December 2024. We requested the mannequin to calculate the worth of the portfolio on the present date.

To perform this process, the mannequin must pull Magazine 7 value data for the primary day of every month, break up the month-to-month funding evenly throughout the shares ($20 per inventory), sum them up and calculate the portfolio worth in line with the worth of the shares on the present date.

On this process, each fashions failed. o1 returned a listing of inventory costs for January 2024 and January 2025 together with a formulation to calculate the portfolio worth. Nonetheless, it didn’t calculate the right values and mainly stated that there can be no ROI. However, R1 made the error of solely investing in January 2024 and calculating the returns for January 2025.

Nonetheless, what was attention-grabbing was the fashions’ reasoning course of. Whereas o1 didn’t present a lot particulars on the way it had reached its outcomes, R1’s reasoning traced confirmed that it didn’t have the right data as a result of Perplexity’s retrieval engine had didn’t get hold of the month-to-month knowledge for inventory costs (many retrieval-augmented era functions fail not due to the mannequin lack of skills however due to unhealthy retrieval). This proved to be an essential little bit of suggestions that led us to the following experiment.

Reasoning over file content material

We determined to run the identical experiment as earlier than, however as an alternative of prompting the mannequin to retrieve the data from the net, we determined to offer it in a textual content file. For this, we copy-pasted inventory month-to-month knowledge for every inventory from Yahoo! Finance right into a textual content file and gave it to the mannequin. The file contained the identify of every inventory plus the HTML desk that contained the worth for the primary day of every month from January to December 2024 and the final recorded value. The information was not cleaned to cut back the handbook effort and check whether or not the mannequin may choose the best elements from the information.

Once more, each fashions failed to offer the best reply. o1 appeared to have extracted the information from the file, however advised the calculation be executed manually in a device like Excel. The reasoning hint was very obscure and didn’t comprise any helpful data to troubleshoot the mannequin. R1 additionally failed and didn’t present a solution, however the reasoning hint contained a variety of helpful data.

For instance, it was clear that the mannequin had accurately parsed the HTML knowledge for every inventory and was in a position to extract the right data. It had additionally been in a position to do the month-by-month calculation of investments, sum them and calculate the ultimate worth in line with the most recent inventory value within the desk. Nonetheless, that last worth remained in its reasoning chain and didn’t make it into the ultimate reply. The mannequin had additionally been confounded by a row within the Nvidia chart that had marked the corporate’s 10:1 inventory break up on June 10, 2024, and ended up miscalculating the ultimate worth of the portfolio. 

Once more, the true differentiator was not the consequence itself, however the potential to research how the mannequin arrived at its response. On this case, R1 supplied us with a greater expertise, permitting us to grasp the mannequin’s limitations and the way we will reformulate our immediate and format our knowledge to get higher outcomes sooner or later.

Evaluating knowledge over the net

One other experiment we carried out required the mannequin to check the stats of 4 main NBA facilities and decide which one had the perfect enchancment in discipline objective share (FG%) from the 2022/2023 to the 2023/2024 seasons. This process required the mannequin to do multi-step reasoning over totally different knowledge factors. The catch within the immediate was that it included Victor Wembanyama, who simply entered the league as a rookie in 2023.

The retrieval for this immediate was a lot simpler, since participant stats are extensively reported on the net and are normally included of their Wikipedia and NBA profiles. Each fashions answered accurately (it’s Giannis in case you have been curious), though relying on the sources they used, their figures have been a bit totally different. Nonetheless, they didn’t understand that Wemby didn’t qualify for the comparability and gathered different stats from his time within the European league.

In its reply, R1 supplied a greater breakdown of the outcomes with a comparability desk together with hyperlinks to the sources it used for its reply. The added context enabled us to right the immediate. After we modified the immediate specifying that we have been on the lookout for FG% from NBA seasons, the mannequin accurately dominated out Wemby from the outcomes.

Closing verdict

Reasoning fashions are highly effective instruments, however nonetheless have a methods to go earlier than they are often totally trusted with duties, particularly as different elements of enormous language mannequin (LLM) functions proceed to evolve. From our experiments, each o1 and R1 can nonetheless make fundamental errors. Regardless of exhibiting spectacular outcomes, they nonetheless want a little bit of handholding to offer correct outcomes.

Ideally, a reasoning mannequin ought to have the ability to clarify to the consumer when it lacks data for the duty. Alternatively, the reasoning hint of the mannequin ought to have the ability to information customers to raised perceive errors and proper their prompts to extend the accuracy and stability of the mannequin’s responses. On this regard, R1 had the higher hand. Hopefully, future reasoning fashions, together with OpenAI’s upcoming o3 collection, will present customers with extra visibility and management.

Each day insights on enterprise use instances with VB Each day

If you wish to impress your boss, VB Each day has you coated. We provide the inside scoop on what firms are doing with generative AI, from regulatory shifts to sensible deployments, so you possibly can share insights for optimum ROI.

Learn our Privateness Coverage

Thanks for subscribing. Take a look at extra VB newsletters right here.

An error occured.