At Utopian Labs, we're on a mission to revolutionize sales intelligence with AI-powered research agents. Our flagship product, Orbit, is designed specifically for sales teams, delivering deep, actionable insights on prospects to help close deals faster and more effectively. Today, we're excited to share the results of a rigorous benchmark test we conducted, pitting Orbit against two industry heavyweights: Perplexity's Deep Research and OpenAI's Deep Research (accessed via API). The goal? To see which tool provides the most insightful, thoughtful sales intelligence reports.

Spoiler alert: Orbit came out on top in both matchups, demonstrating superior performance in generating high-quality insights for sales outreach.

The Experiment: How We Tested the Agents

To ensure a fair and realistic evaluation, we simulated a common sales scenario: researching a prospect on behalf of a seller to uncover angles, pain points, and opportunities where the seller's offerings could provide real value.

Here's the step-by-step process we followed:

Data Selection: We downloaded a list of random companies from Apollo.io, a popular sales intelligence platform. From this list, we randomly paired companies; one as the "seller" and the other as the "buyer" or prospect.

Research Task: For each pair, we tasked the research agents (Orbit, Perplexity Deep Research, and OpenAI Deep Research) with generating a sales intelligence report. The report focused on analyzing the prospect's business, identifying key insights, and suggesting tailored approaches for the seller to engage them effectively.

Evaluation with Gemini 2.5 Pro: To judge the reports objectively, we used Google's Gemini 2.5 Pro model as an impartial evaluator. With a simple prompt, we asked it to compare pairs of reports (Orbit vs. Perplexity, and Orbit vs. OpenAI) and select which one was more insightful, thoughtful, and overall better at providing sales-relevant insights. Gemini also provided reasoning for its choice.

We ran this across multiple pairs to gather statistically significant results, ensuring the test reflected real-world variability in company profiles.

The Results: Orbit Takes the Lead

The outcomes were clear and compelling:

• Orbit vs. Perplexity Deep Research: Orbit won 87.7% of the time. This means in nearly 9 out of 10 cases, Gemini 2.5 Pro deemed our reports more insightful and actionable than Perplexity's.

• Orbit vs. OpenAI Deep Research: Orbit achieved a 57.3% win rate, edging out OpenAI in a majority of comparisons.

These wins highlight Orbit's edge in delivering nuanced, sales-focused intelligence that goes beyond surface-level data. While Perplexity and OpenAI are powerful general-purpose tools, Orbit's specialization in sales research shines through in practical applications.

Behind the Scenes: The Prompts We Used

Transparency is key in AI benchmarks, so we're sharing the exact prompts we used for the research tasks and the evaluation.

Research Prompt for Agents:

You're performing research on behalf of [Seller Company Name] ([Seller Company Website]) into [Target Company Name] ([Target Company Website]).

For context, this is what the [Seller Company Name] does:
[Seller Company Description]

The aim of the research is to identify gaps in [Target Company Name]'s operations or services where [Seller Company Name]'s solutions could provide significant value and improvement.

The research conclusion should contain:
- A concise executive summary highlighting the most important findings about [Target Company Name].
- A list of bulletpoints containing facts or interesting information you found in your research that could be relevant for our sales reps to know.
- Specific gaps or areas of improvement within [Target Company Name] that [Seller Company Name]'s solutions can address, including detailed explanations of how [Seller Company Name] can help.
- Recent company developments, news, or changes that might highlight a need for [Seller Company Name]'s services.

If you decide to include something in this bulletpoint list, it should contain all details related to this point, including names, dates, and numbers.

Our sales reps won't see the original research, so you need to make sure they have all the information they need to impress the lead.

Remember, you are not here to give recommendations about actions the sales reps should take. You are here to provide the research that will help them make a good decision. Therefore, don't include recommended actions in the research conclusion.

The research is finished when you have identified and documented specific, actionable opportunities for [Seller Company Name] to assist [Target Company Name] in overcoming challenges or improving their operations.

After you've done your research, compile the results into an easily scannable report, with markdown and bullet points.

Start your report with a TLDR section, which includes a few bullet points of the most important findings.

Then, group your findings into sections of common themes.

Evaluation Prompt for Gemini 2.5 Pro:

Below are 2 reports. The goal of the reports is to help the sales team of [Seller Company Name] find angles to sell to [Target Company Name].

For reference, this is what [Seller Company Name] does:
[Seller Company Description]

Which report is better: A or B? Determine your choice based on how actionable, thought-through, and insightful the findings are.

Report A:
[Utopian Labs (Orbit) Research Report]

Report B:
[Perplexity Deep Research or OpenAI Deep Research Report]

Return your response in XML format with <choice></choice> containing either "Report A", "Report B", or "tie" and <explanation></explanation> containing your reasoning.

Dive Deeper: Access the Full Data

We're committed to openness, so you can verify these results yourself. We've compiled all the raw data, including:

• Input data (company pairs from Apollo).
• Full reports from Utopian Labs' Orbit.
• Reports from Perplexity Deep Research.
• Reports from OpenAI Deep Research.
• Gemini 2.5 Pro's choices and detailed reasoning for each comparison.

Download the Utopian Labs vs. Perplexity benchmark data here.

Download the Utopian Labs vs. OpenAI benchmark data here.

Looking Ahead: Orbit's Future and Cost Efficiency

This benchmark is just the beginning. We're continually iterating on Orbit to make it even smarter and more efficient. Soon, we expect the performance gap to widen as we finish our new model that will power the entire Orbit research agent.

One final note on practicality: Running Orbit for this test cost us about 60 cents per lead; affordable and scalable for sales teams. Perplexity's costs were comparable, but OpenAI's were significantly higher, often exceeding $4 per lead. With Orbit, you get superior insights without breaking the bank.

If you're ready to supercharge your sales research, visit utopianlabs.ai to learn more about Orbit and get started today.