Pendless vs. Perplexity Comet Browser Agent

Inside Our November 2025 Benchmark

Over the past year, we’ve been building Pendless with a simple goal: create a browser automation engine that works the way people expect software to work. Fast, accurate, stable, and able to complete real business workflows from start to finish. As we approached launch, we wanted to validate our progress against one of the most well known browser agents in the market: Perplexity’s Comet.

This wasn’t a casual test. We designed a structured benchmark that mirrored real business conditions and common browser workflows. Both tools were evaluated using the same set of 34 prompts across identical websites, environments, and task descriptions. The goal was straightforward: measure accuracy, completeness, speed, and reliability under controlled, repeatable conditions.

What We Tested

The dataset represented everyday browser work: entering data into forms, moving information between systems, navigating interfaces, interacting with tables, and performing multi-step workflows. Everything was designed to reflect the type of tasks businesses run hundreds or thousands of times each week.

For each prompt, we scored four key measures:

Precision The percentage of actions executed correctly from the actions attempted.

Recall How many of the required workflow steps were successfully completed.

Execution Time The time from start to finish.

General Errors Crashes, stuck states, missed steps, or incorrect task completions.

These metrics matter because they reflect real production behavior. High precision means fewer mistakes. High recall means workflows reach the finish line. Low error rates reduce human oversight. Faster execution increases throughput and ROI.

The Results

Pendless outperformed Comet across every major dimension.

Average execution time Pendless averaged 23.4 seconds per task. Comet averaged 2 minutes and 13 seconds. Pendless was roughly 480 percent faster.

Precision Pendless achieved 97.2 percent. Comet reached 93.3 percent.

Recall

Pendless delivered 97.9 percent. Comet completed 93.6 percent.

General errors

Pendless recorded zero failures. Comet recorded six.

Speed, accuracy, completeness, and stability all moved in the same direction. Pendless didn’t just edge ahead. It demonstrated measurable, material advantages in the areas that affect real productivity.

Why Pendless Performed Better

While we don’t disclose proprietary implementation details, several high-level architectural choices contributed to these outcomes.

Pendless uses a lightweight execution engine designed specifically for browser tasks. It aligns directly with real page structures visible at each moment rather than relying on vision models or in-memory replication of the user actions. It plans actions deterministically to reduce misclicks and hallucinated behaviors. Our page snapshot conditioning and action cycle timing reduce delays and keep workflows moving smoothly. Our fallback logic helps the engine avoid dead ends and recover gracefully from unexpected states.

Another key factor is orientation. Pendless is built for robotic consistency rather than open-ended agentic reasoning. That means fewer speculative actions, fewer incorrect assumptions about the page, and a tighter coupling between what the engine sees and what it does. This robotic posture produces steadier action sequences, higher accuracy, and a level of predictability that agentic systems often struggle to match.

These choices compound in practice. They show up as faster runs, cleaner execution, fewer skipped steps, and zero blocked workflows.

What This Means for Businesses

For teams handling repetitive browser work, the difference between a tool that completes tasks correctly on the first pass and a tool that struggles can be enormous. Faster agents reduce queue backlogs. Higher accuracy protects data integrity. Fewer errors mean less manual cleanup. And when workflows complete end to end, businesses unlock real scale.

This benchmark reinforces what we’ve believed all year: automation should work reliably, predictably, and without drama. The tool should feel like a partner, not a project. It should free humans for meaningful work by taking care of the monotony.

Looking Ahead

Pendless is preparing for launch with an additional engine optimization sprint, so even these results represent a baseline. As we continue refining the engine, we expect further gains in speed and consistency.

This benchmark is a milestone, but not a finish line. It confirms that the architecture is sound, the foundations are strong, and the tool is ready for broad deployment across SMB and enterprise environments.

If your team relies on browser workflows, Pendless is built for you. Let’s automate the monotony and build something better together.

Download the full benchmarking report.