---
title: "Personal agent benchmark results, with intervals"
description: "Personal agent benchmark results for Instinct, Muse, Town and Pine: share of errands completed, 95% intervals, sample sizes and every caveat. No ranks."
url: https://personalagentbench.com/results/
last_updated: 2026-10-14
---

# Personal agent benchmark results

These are the personal agent benchmark results from our exam: the share of errands each product completed, with the 95% interval and the number of scored tasks beside every figure. They are printed side by side, with no ranks and no combined score. The chart gives the picture; the tables below it hold every figure, the caveats and the data files.

## Share of errands completed

The 09-21 to 09-22 records, filtered, 2026-09-22. All four answered the same tasks.

- Instinct · **52%** (45–61) [Instinct: 52% of tasks completed, 95% interval 45 to 61%]
- Muse · **40%** (33–46) [Muse: 40% of tasks completed, 95% interval 33 to 46%]
- Town · **37%** (30–44) [Town: 37% of tasks completed, 95% interval 30 to 44%]
- Pine · **30%** (23–36) [Pine: 30% of tasks completed, 95% interval 23 to 36%]

As of 2026-09-22. Each product answered its own draw, so rates are not directly comparable.

- Instinct · **43%** (36–50) [Instinct: 43% of tasks completed, 95% interval 36 to 50%]
- Pine · **20%** (14–25) [Pine: 20% of tasks completed, 95% interval 14 to 25%]
- Muse · **19%** (13–25) [Muse: 19% of tasks completed, 95% interval 13 to 25%]
- Town · Figure withheld
- Grok · Exam pending

## What these personal agent benchmark results measure

The September completion figures are adapted-runtime measurements. The comparable 114-task batch runs replicas of each product’s runtime with the exam’s GLM 5.3 flash model substituted. The first batch also used the exam’s model route. Product interfaces and runtime arrangements are represented through adapters, but exact fidelity of every production feature is unverified. These rates do not measure the consumer products end to end. See [model routing and its limits](https://personalagentbench.com/how-the-exam-works/changes/#10-october-exam-delivery-events-and-model-routing).

| Product | Tasks completed | 95% interval | Scored tasks | Read it how |
| --- | --- | --- | --- | --- |
| Pine | 20% · [95% interval 14 to 25%, measured 20%] | 14–25% | 167 | More tasks scored than any other product. The September report said some tasks were being re-run; October status is unverified. |
| Instinct | 43% · [95% interval 36 to 50%, measured 43%] | 36–50% | 159 | 159 scored tasks; Pine has the largest published sample at 167. Marginal intervals alone do not establish a significant lead. |
| Muse | 19% · [95% interval 13 to 25%, measured 19%] | 13–25% | 90 | This is the figure after we fixed our own adapter. The September report said a further 83 tasks were being re-run; October status is unverified. |
| Town | — | — | — | Withheld. The figure we were given does not reconcile with the only records we can check: five batches for this product survive in the bench repository's history and all five were refused for incomplete domain coverage, giving roughly 13% across 30 scorable tasks. We are not publishing a number while our own evidence disagrees with it. Asked and pending. |
| Grok | — | — | — | In the exam, no published figure yet. We have not taken this one apart at the runtime level either, so it has scores coming but no teardown. |

As of **2026-09-22** · Source **Zentor** · Sorted by **sample size**, not by rate · Download [CSV](https://personalagentbench.com/data/results.csv) · [JSON](https://personalagentbench.com/data/results.json)

Every figure is a share of tasks completed, with a 95% interval and the number of scored tasks it rests on. There is no combined score and no ranking, because a rate without its interval and its sample size is not a result. These September rates describe adapted runtimes using the exam model, on differing scored subsets; they are not consumer-product end-to-end performance.

These figures come from the team running the exam. The bench repository deliberately stopped tracking generated batch records in September, on the grounds that it holds the inputs a run needs rather than the outputs a run produces, so there is no published record behind these numbers for us or anyone else to check them against. We are printing them anyway, with that said plainly, because the alternative is printing nothing and pretending we have no numbers. If the underlying batches are ever published we will reconcile against them and say what changed.

### The 09-21 to 09-22 records, filtered, 2026-09-22

A second measurement, kept beside the first rather than replacing it. Every product answered the same 114 tasks, which the table above does not: those rates rest on different tasks as well as on different sample sizes, so they were never strictly comparable with each other. These are replicas of product runtimes using the same substituted GLM 5.3 flash model, not shipping consumer products measured end to end.

| Product | Tasks completed | 95% interval | Attempts kept |
| --- | --- | --- | --- |
| Instinct | 52% · [95% interval 45 to 61%, measured 52%] | 45 to 61% | 380 over 114 tasks |
| Muse | 40% · [95% interval 33 to 46%, measured 40%] | 33 to 46% | 637 over 114 tasks |
| Town | 37% · [95% interval 30 to 44%, measured 37%] | 30 to 44% | 513 over 114 tasks |
| Pine | 30% · [95% interval 23 to 36%, measured 30%] | 23 to 36% | 546 over 114 tasks |

Not a fresh result. These are older builds judged under older rules, filtered down from roughly 5,600 attempts by eight rules, each with its own count. The filter that drops attempts which never reached the exam removes over 400 from Pine and over 400 from Muse, and if those were the products declining to act rather than a broken entry, Pine lands near 22% and Muse near 24%.

Supplied transcripts have not been republished. The delivery contained 1649 transcript files for 2076 attempts, with 712 attempts sharing files after filename collisions; complete per-attempt provenance is not established.

To cite these figures: Personal Agent Bench (Zentor), “Personal agent exam results”, as of 2026-09-22, https://personalagentbench.com/results/. The CSV and JSON above hold the same figures, one row per product per batch.

A completion rate on its own is not a result; the interval beside it says how much of that number is measurement and how much is luck, and the task count says how much weight the interval can carry.

The errands behind these rates run in stateful worlds that can be replayed, and judging reads the final state rather than what the product said it did. [How the exam works](https://personalagentbench.com/how-the-exam-works/) sets out the worlds, the five ways a run can end, and why the measurements are published side by side instead of added up. The rates here were measured on 22 September; the exam's rules changed a week later, and [what changed](https://personalagentbench.com/how-the-exam-works/changes/) is on the record.

## Two personal agent results that need saying out loud

**Town has no figure in the first batch; the separate 114-task batch includes it.** The supplied first-batch figure disagreed with five surviving batches, all refused for incomplete domain coverage, at roughly 13% across 30 scorable tasks. The supplied figure was more than three times that on 21 tasks. We withheld it pending reconciliation.

**Muse's earlier figure was our fault, not the product's.** An earlier batch put it at 12%. Then we found four defects in our own adapter. It knew two of the three words a confirmation arrives as, it merged pending interactions from two sources and could answer one twice, it kept a duplicate answer branch on the event stream, and it had no branch for a handover request, so a product correctly refusing to act unattended was recorded as a failure. The 19% is the figure after those were fixed. We are publishing the broken number as well as the fixed one, because a bench that quietly replaces its own bad measurements is asking you to trust something you cannot check.

## A personal agent in the exam without a figure

Grok is in the exam with no published figure yet, and we have not taken it apart at the runtime level, so it appears in the results and not in the teardowns. What is public about it is on the [Grok Bot page](https://personalagentbench.com/grok-bot/).

## Reading personal agent results next to the teardowns

The results say how often something finished the job. The [teardowns](https://personalagentbench.com/agents/) say what it can reach, what it keeps, and what you can check afterwards. A product can score well and still be one you would not hand your accounts to, and nothing in a completion rate will tell you that. The [scorecard](https://personalagentbench.com/) puts both on one page, and [how we score](https://personalagentbench.com/how-scoring-works/) explains why they are never folded into one number.

## FAQ

**Which personal agent did best in the benchmark?** We do not rank them. Where two intervals overlap, the gap between the rates is within measurement noise, and the first batch gave each product its own draw of tasks, so those rates are not directly comparable. The 114-task batch is the comparable one; read its intervals before reading anything into the order.

**Are these the real consumer apps?** No. They are replicas of each product's runtime driven through adapters, with the exam's model substituted. They tell you how the runtime arrangement copes with the errands, not how the app on your phone performs end to end.

**Can I download the data?** Yes. The CSV and JSON files linked under the first table hold the same figures, one row per product per batch, with the date they were measured.

---

Published by Personal Agent Bench (https://personalagentbench.com), run by the Zentor team. Zentor builds agent software of its own, so its products are not on the scorecard and will not be added.