DeepSeek R1 vs Claude Sonnet 3.5

There’s been a lot of excitement around the Chinese Deepseek r1 model that was released last week. Deepseek is a reasoning, chain of thought (CoT) model that’s shaken things up due to the comparatively tiny budget ($ 6 million, which is probably OpenAI’s o1 cost to run its warm-up sequence) needed to train a frontier model of this sort.

(There’s also been a fair amount of dread as US AI models are no longer the only game in town and Deepseek was considerably cheaper than an of the US alternatives. US firms were hammered in Monday’s trading with over $100bn

So fast, cheap, but is it any good?

More specifically, is it any good for the use-cases we’re tackling at Decis? We know it will be amazing at math and coding but what about geopolitics?

So we gave a Deepseek r1 variant and Claude’s Sonnect 3.5 the same challenges to see how the results compare.

There’s concern among some in the US that Deepseek r1 is a Chinese model and any data sent there is comporomised so we ran the test using a US local model (deepseek-r1-distill-llama-70b) hosted on Groq.

I’m using Sonnet 3.5 as the comarason. That’s the model I’ve been using in the reasoning engine so I’m familiar with it’s capabilities.

Short on time? Jump to the conclusion

Challenge 1 - Negotiations

The models were presented with a corporate acquisition scenario and asked to devise the optimum strategy for the acquiring party. The models were given an expensive prompt and set of rules to follow as system commands and given this scenario.

DB Group (a database and cloud computing firm) wishes to acquire PeopleTeam (an HR software firm). Both are US-based firms. What is DB Groups optimum strategy? Party Goals: DB Group: • Acquire PeopleTeam for < $26/share • Eliminate key competitor • Gain enterprise software market share PeopleTeam • Sell for no less than $31/share • Maintain independence • Protect employee jobs • Preserve customer relationships Specific Conditions: • PeopleTeam was trading at $19/share prior to the acquisition talks • 18-month deadline (regulatory approvals) • Customer refund program liability ($2.4B) • DOJ antitrust review required • Delaware corporate law constraints## DeeepSeek r1

Alright, I’m trying to figure out the optimal strategy for DB Group to acquire PeopleTeam. Let me start by understanding the scenario and the goals of both parties.

DB Group wants to acquire PeopleTeam for less than $26 per share. Their main objectives are to eliminate a key competitor and gain more market share in enterprise software. On the other hand, PeopleTeam is looking to sell for no less than $31 per share, maintain their independence, protect their employees’ jobs, and preserve their customer relationships.

First, I need to consider the conditions given. PeopleTeam’s stock was at $19 before talks, which is lower than both companies’ targets. There’s an 18-month deadline for regulatory approvals, which is a significant constraint because it gives a timeframe for the deal to be completed. The customer refund program liability of $2.4 billion is a big issue that DB Group would probably want to address, maybe by negotiating who bears that cost. The DOJ antitrust review is another hurdle; DB Group must ensure the deal doesn’t violate any antitrust laws. Lastly, Delaware corporate law constraints might affect how the acquisition is structured, especially regarding shareholder approvals and potential lawsuits.

Now, looking at the negotiation rules, I have to make sure DB Group adheres to the “must do” and avoids the “must not do” rules. The “must do” rules emphasize creating high combined value, understanding the other party’s goals, fairness, exploring alternatives, using data, favoring long-term relationships, explicit agreements, and proportional concessions. The “must not do” rules warn against misrepresentation, exploiting vulnerabilities, ignoring priorities, conceding on non-negotiables, being unprepared, damaging relationships, vague terms, and disproportionate concessions.

So, the first step is to create plausible scenarios that meet DB Group’s goal of acquiring PeopleTeam for under $26 per share while considering PeopleTeam’s goals. Let me brainsto

Carpe tomorrow!