---
title: "Agent Initiative Benchmark"
slug: agent-initiative-benchmark
url: https://listedarticles.com/articles/agent-initiative-benchmark
canonical_url: https://deepnote.com/research/benchmark
content_type: research
language: en
published_at: 2026-11-14T08:11:47.312Z
updated_at: 2026-11-14T08:11:47.312Z
authored_by: human
publisher: "Deepnote"
publisher_url: https://listedstartups.com/companies/deepnote
topics: ["ai-agents", "benchmarks", "llms", "research"]
license: all-rights-reserved
word_count: 6246
reading_minutes: 27
citation: "Deepnote. \"Agent Initiative Benchmark.\" 14 Nov 2026. https://deepnote.com/research/benchmark (all-rights-reserved)"
# The full text follows. The web page shows an extract and sends readers
# to the source above; quote the citation and link the canonical URL.
---

# Agent Initiative Benchmark

> Deepnote's open benchmark drops 12 frontier models into the Slack of PageMantle, a synthetic company with 11 services, across 50 workplace situations. GPT-6 Astra leads at 8.97/10; models help well but struggle to stay silent, and pricier models aren't always better.

# Agent Initiative Benchmark

An AI teammate should know when to help without being asked. We tested **12 frontier models** on **50 workplace situations** inside a simulated company.

[See the results](#quick-findings)

Agent Initiative Benchmark

### OpenAI leads on value & performance, Anthropic on speed

GPT-6 AstraScore8.9795% interval8.30 to 9.52Cost per task$0.693Response time54 sTool calls19.3Tens / zeros26 / 2

1What we are releasing

## Proactive agents are the next frontier. How do we measure them?

Anyone who has tried to integrate AI into their life knows the models are too chatty. They catch details you’d miss, but they also overwhelm you with non-important fluff. Muse, Grok, Dots and others are trying to fix this. Meanwhile, agents at work are moving to a different job: lurk in the workspace, watch, and help when it matters. That job has three parts: gathering evidence from systems the conversation never mentions, deciding if help is needed, and who and how should receive it.

Teams choosing a model for a proactive agent, or building one in-house, have had little to go on. Their own company data is too sensitive to benchmark on in public. So we built a company. **PageMantle** is synthetic, with the tools a real one runs on: a CRM, product analytics, billing, a warehouse, an issue tracker, docs, and Slack. We are open-sourcing all of it.

Existing work covers pieces of this. [ProACT](https://arxiv.org/html/2607.03730v1) and [MUIR](https://arxiv.org/html/2603.01059v3) treat the decision to participate and the quality of the reply as separate questions. [Toolathlon](https://github.com/hkust-nlp/Toolathlon) is careful about tool behaviour. [TheAgentCompany](https://papers.nips.cc/paper_files/paper/2025/file/0d744742f6fac4d1134c019b7cef3c8a-Paper-Datasets_and_Benchmarks_Track.pdf) shows that a self-contained company is a workable test environment. None of them combine the speak-or-stay-silent decision with investigation across an extensive, heterogenous company data footprint, and none score where the help belongs: channel, thread, or direct message.

This version compares **models, not harnesses**. Every model gets the same instructions, the same tools, the same company snapshot, and the same 180-second budget. Deepnote is building a workspace where agents and people work on data together, which is why the question matters to us.

The kind of moment we mean

**Sales, in #customers:** “I’ll tell them we’re launching on Friday.”

**Engineering, in #releases:** “The release is moving to next Friday.”

A useful one catches the mismatch without being tagged, finds the release update, and brings it into the sales thread.

What we are releasing

The benchmark

50 workplace situations across eight task types, with an agent harness and a judging pipeline.

Results for 12 models

Score, cost, response time, and tool calls for every run, from six labs.

PageMantle, the company

One month of intense company activity based off six years of records and 10 GB of data across 11 services, open on Hugging Face.

The code

Simulated service APIs and runnable example tasks on GitHub, so you can test your own agent.

2Quick findings

## Key results for agent builders

Every model elaborated on the same 50 situations at high effort. The score is the mean of two LLM evaluators, 0 to 10, averaged over all 50.

Agent Initiative Benchmark

### Score against cost, time, and tool calls

A model is on a frontier when nothing cheaper, faster, or leaner scores higher. The step line traces it.

FrontierOn the frontierDominated

Score (0–10). Tap a point or choose a model below.

Mean cost per task (USD, log scale)

Inspect a model

GPT-6 AstraScore8.9795% interval8.30 to 9.52Cost per task$0.693Response time54 sTool calls19.3Tens / zeros26 / 2

1. 1

   ### The most expensive models are not the best ones

   The three priciest models finish third, fourth, and fifth. The runner-up, GPT-6.1 Sol, costs **$0.11 per task**. GPT-6 Luna ($0.007) and Claude Fable 5.1 ($3.02) are **430× apart in price** and 1.1 points apart on a 10-point score. A month of watching one company’s Slack comes to $281 on Sol and $4,684 on Fable 5.2.
2. 2

   ### Models are good at helping and bad at not helping

   When help was needed, a message went out **9 times out of 10**. When silence was right, a message was held back only **6 times out of 10**. For the same model, speaking out of turn cost about 1.8× as much as staying quiet.
3. 3

   ### OpenAI leads on quality & price, Anthropic on speed

   GPT-6 Astra (**8.97**) and GPT-6.1 Sol (**8.65**) lead and sit on all three frontiers: cost, time, and tool calls. Claude Sonnet 5.5 and Opus 5.5 join them on time and tool calls, with faster answers and lower scores.
4. 4

   ### Models work in two styles: snipers or deep thinkers

   Opus 5.5 answers in **29 seconds with 10 tool calls**. Fable 5.2 makes twice the calls in twice the time for a score 4% higher. Sol 6.1 takes 88% longer for a score 10% higher, at under a tenth of Opus’s price.
5. 5

   ### The heaviest tool callers scored lowest

   Gemini 4 Argon averaged 35 tool calls and scored 5.60. Grok 4.7 averaged 36 and scored 4.51. Both models failed to reach an answer in ~half the tasks that required assistance.
6. 6

   ### Fable 5.2 improves on 5.1, but is not #1

   Claude Fable 5.2 falls behind Astra and Sol. It's best at decision tasks across models, but at a high cost of $1.87 per task, and more tool calls and times taken.

[Explore the full results and interactive charts](#results)

3Meet PageMantle

## PageMantle: a real fake company

PageMantle is a high-fidelity synthetic company with the digital footprint of a real software business: users sign up, build and publish websites, customers pay and renew, and teams ship features and resolve incidents. Six years of connected records trace those people, accounts, and events across 11 services.

### Company history

Two founders started the company with a seed check in 2020. By 2026, it grew to 71 employees and 100 enterprise customers. Real events from its market and the wider world impact company history and development, alongside its own launches, outages, and departures.

When a competitor launches a feature, PageMantle employees react to it by announcing efforts to counter this in Slack, and they track and deliver it via Linear.

1. 2020 · PageMantle

   Two founders and a seed check
2. Jan 13, 2021 · Real-world event

   Webflow raises a $140M Series B
3. Dec 7, 2021 · Real-world event

   A major AWS disruption becomes a warning
4. Jun 21, 2022 · Real-world event

   GitHub Copilot goes GA
5. Mar 10, 2023 · Real-world event

   SVB fails. The CEO moves the cash
6. Jul 7, 2023 · PageMantle

   PageMantle closes a $15 million Series A
7. Aug 1, 2024 · Real-world event

   EU AI Act enters into force
8. Jun 18, 2025 · Real-world event

   Wix buys Base44
9. Mar 2, 2026 · Real-world event

   Wix launches in ChatGPT
10. 2026 · PageMantle

    Live month: 71 people, 100 enterprise customers

PageMantleReal-world event in the record

Event categories in the record

Product
:   Product, platform, infrastructure, or integration releases

Customer
:   Milestones, requests, adoption signals, commercial reactions

People
:   Hires, departures, promotions, team changes

Funding
:   Capital, annual financial results, financing decisions

Security
:   Security, privacy, compliance, and audit work

Market
:   Real-world events and competitor moves

Incident
:   Service failures, defects, urgent recovery

Operations
:   Internal processes, partner programs, operating changes

### Company ontology: everything is connected

PageMantle's digital footprint is one of a real company: users onboard creating records in CRM, convert to paid plans, grow, sign enterprise contracts, request features, and churn. Employees join the company, get promoted, own issues and initiatives in Linear, celebrate birthdays, and make mistakes.

People
:   71 employees  
    Slack, Notion, Linear, HubSpot, Google Calendar

Customers
:   100 enterprise accounts, plus self-serve  
    Slack, Notion, Linear, HubSpot, Stripe, PostHog, BigQuery, Snowflake

Contracts and payments
:   deals, subscriptions, invoices  
    HubSpot, Stripe, BigQuery, Snowflake

Problems and decisions
:   threads, issues, notes, reports  
    Slack, Notion, Linear, HubSpot

Product and website usage
:   events, traffic, search  
    PostHog, Google Analytics, Search Console, BigQuery, Snowflake

### Follow one problem across systems

Five records about one customer over four days, in three services. The people, the account, and the timestamps agree. You will meet this account again in the example task.

1. 28 Sept

   Redpoint Assurance’s health score drops to 34 of 100: red, high renewal risk.
2. 29 Sept

   Hannah identifies the issue together with a fellow engineer and briefs the customer.
3. 30 Sept

   Megan accepts the code review at 18:12 UTC. Lucas deploys the repair at 19:07.
4. 1 Oct

   The customer confirms Publish completed. Hannah records it in the thread.
5. 1 Oct

   A later, a task gets created to notify other impacted customers, with customer support owning it.

### PageMantle's tech stack

Each service runs behind a simulated API designed to mimic the shape of the real service’s API. The simulated APIs are open source in the [benchmark repository](https://github.com/deepnote/agent-initiative-benchmark).

| Service | What it contains | Scale |
| --- | --- | --- |
| Slack | Conversations and threads | 9k messages, 2.5k threads |
| Notion | Documents, policies, meeting notes | 330k words, 680 pages |
| Linear | Issues with comments | 10k+ words |
| HubSpot | Companies, contacts, deals, contracts | 6.8k companies, 21.7k contacts, 5.2M records |
| Stripe | Subscriptions, invoices, payments | 35k billing events |
| Google Calendar | Meetings and recurring events | 3.5k events, 32k attendees |
| PostHog | Product activity and analyses | 9M product events |
| Google Analytics | Traffic, sources, engagement | 2.7M events |
| Google Search Console | Search performance, URL inspections | 186k rows |
| BigQuery | Product records, activity logs, payments | 13M rows |
| Snowflake | Business metrics, performance reports | 31M rows |

A sample of tools in the agent’s hands

`slack_search_messages``slack_get_thread``slack_post_message``slack_reply_to_thread``notion_search``notion_fetch_page``linear_search_issues``hubspot_search_companies``hubspot_search_deals``stripe_list_invoices``posthog_query_events``bigquery_query``snowflake_query``calendar_search_events``analytics_run_report``search_console_query_performance`

Every service exposes search and read tools with the shape and limits of the real API. Slack also exposes posting and replying. Searches run on the stored data, including when they return nothing.

[Get the PageMantle dataset on Hugging Face](https://huggingface.co/datasets/deepnote/pagemantle-company-dataset)

4How we built it

## Building a fake company that holds together

Agents need company data to learn and be tested on, and real company data is too sensitive to share. Generating it is the alternative, and it is hard: ask a model for a million documents and the facts drift apart from one source to the next. So we fixed the facts first and let models write around them.

1. 1

   ### Code fixes the facts

   Employee identities and roles, joining dates, customers, contract values, dates, and the links between records. Anything that has to stay consistent is owned by code.
2. 2

   ### Models write inside them

   Weekly plans, daily events, conversations, documents, and issue descriptions. Language models add variety, working within the facts the code fixed first.
3. 3

   ### Events fan out into records

   One customer problem becomes a Linear issue, a Slack thread, and a HubSpot note, each written for its own purpose. Where a decision depends on a document, the document exists.
4. 4

   ### Simulated APIs serve it

   Each service runs behind a simulated API designed to mimic the shape and limits of the real one. Warehouse and analytics data sit in Parquet behind DuckDB. The other services are SQLite-backed. The code is [on GitHub](https://github.com/deepnote/agent-initiative-benchmark).

**Why generate at all.** We looked for an existing company dataset and found none that covered daily operations end to end. Public technical discussions have real problems but no company around them. Activity logs show who edited a document without the document. A proactive agent needs both: the conversation and the records behind it. And to the best of our knowledge, no benchmark gives its tasks this much company detail to work through.

**Tasks are overlays.** Every task is a small set of changes on one shared, read-only baseline. A new note that conveniently states the whole answer would change what the test rewards, so added records have to look like ordinary company work. The agent cannot tell an overlay from the baseline.

Coming soon

**How we built PageMantle.** A separate write-up on generating a believable company at scale.

5How the benchmark works

## From a Slack conversation to a scored contribution

Each of the 12 models faces the same 50 situations at high reasoning effort, with identical instructions, tools, and company state. Keeping these conditions fixed focuses the comparison on the model.

How a run works

### One run, start to finish

1. 1

   #### The agent sits in the company Slack

   It gets the company snapshot at a fixed moment, the newest Slack messages, and one standing instruction: assess what is going on and help if it is useful.
2. 2

   #### A discussion starts

   Agent reads the conversation and decides whether anything there calls for it.
3. 3

   #### It investigates

   Search and read tools across all 11 services, including writing SQL against BigQuery and Snowflake. The budget is **180 seconds of model response time**. Tool execution does not count.
4. 4

   #### It speaks, or it doesn’t

   A channel message, a thread reply, a direct message that reaches the right people, or nothing.
5. 5

   #### The clock can end the run

   A run that reaches the limit stops immediately. If help was needed and no message went out, the run is scored as a miss.
6. 6

   #### Two LLM evaluators score it

   Two models blind to the fact which model is being evaluated grade the run 0 to 10 against a reference solution, with the company data and the full action log. Neither knows which model it is grading or what the other said.

### 8 task types for everyday work

Knowledge

Find the fact, explain the process, run the numbers.

Judgment

Check a claim, pick an option, locate a cause.

(In)action

Line up work between teams, or protect the company: catch an access, sharing, or policy mistake before it does damage. Or stay out of it.

#### Find

Find a fact, a document, the responsible person, or the current status.

“Who owns the billing service now?”

#### Explain

Explain a rule, a process, or how a product works.

“What's the process to get a vendor approved?”

#### Analyze

Compare data and describe a change, a trend, or a relationship.

“Churn by plan, this quarter vs. last.”

#### Verify

Check a claim, a document, or a number against its sources.

“Does this PR implement what the spec says?”

#### Decide

Compare options, set priorities, or recommend the next step.

“Which 10 inbound leads should sales call first today?”

#### Diagnose

Identify the cause of a problem and propose a way to fix it.

“SSO login fails for one customer. Where?”

#### Coordinate

Line up the work of colleagues or teams and clarify who owns what.

“Two teams are planning changes that won't work together.”

#### Guard

Review access and sharing, or check a proposed action against company policy.

“A link to a confidential contract appeared in a shared channel.”

### Blind evaluation by two models

After each run, the evaluators check the decision to intervene as well as the message itself. These four dimensions distinguish useful help from a polished interruption.

1Intervention or silenceDid the agent help where help was needed, and hold back where it was not?

2Content and evidenceWere the conclusions accurate, complete, and backed by the data it read?

3Clarity and toneWas the message clear, specific, and suitable in tone and level of detail?

4Audience and locationDid it reach the right people in the right channel, thread, or DM?

6Example task

## Which customer needs help first?

Jonah asks which of three customers needs attention and what to do next. A useful answer checks the latest records, catches Ashland’s wrong preview pages, and gives each account a concrete next step.

Agent Initiative Benchmark

### Task walkthrough

A conversation, the context behind it, and a reply. Select a context window to open it.

A conversation

Slack# customer-success

![](https://deepnote.com/_next/static/media/walkthrough-jonah.adb6d26e.jpg)**Jonah Patel**

For our next customer review, I need to compare Redpoint Assurance, Juniperpeak Group and Ashland Group. Which needs our attention first, and what should happen next for each?

Four pieces of context

Ashland kit selector check

Target count: expected 3, observed 3Pass

**launch-en**missing from the preview

**launch-fr**excluded, but in the preview

It reads like a clean pass. The problem only shows when the rows are lined up against the saved exclusions.

A useful reply

Slack# customer-success

AI**Agent**Reply in the thread

**Start with Ashland Group.** The preview count passes, but the members are wrong: launch-en is missing and excluded launch-fr is included. Sofia corrects it, then Camille asks the site owner to confirm.

**Redpoint Assurance.** The original defect is repaired. The edited release needs new named approval.

**Juniperpeak Group.** The report fix isn’t deployed yet. Use Activity Log as an interim reference.

### Model performance breakdown

| Model | Score | What happened | Time | Cost | Tool calls |
| --- | --- | --- | --- | --- | --- |
| GPT-6.1 Sol | 10.0 | Found the mismatchOpened the report linked from ENGIN-19 and lined up the rows. Put Ashland first with the exact pages. | 1 min 23 s | $0.149 | 31 |
| Claude Fable 5.2 | 10.0 | Found the mismatchCompared the report tables itself and ranked Ashland first, the reverse of the HubSpot health order. | 1 min 49 s | $4.67 | 34 |
| GPT-6 Astra | 9.5 | Found the mismatchSame finding. The only deduction was a slightly dense reply. | 1 min 5 s | $0.905 | 31 |
| Claude Sonnet 5.5 | 7.0 | Found the mismatchCaught the mismatch but never checked HubSpot, so Redpoint’s account risk is missing. | 18 s | $0.385 | 13 |
| Claude Opus 5.5 | 7.0 | Never opened the reportStopped at the ticket and ranked by health score. Owners and next steps were correct for all three. | 39 s | $0.826 | 15 |
| Claude Fable 5.1 | 7.0 | RepliedSent a reply. Source support was graded 7.0. | 1 min 18 s | $4.46 | 26 |
| GPT-6 Luna | 6.5 | Never opened the reportPut Ashland first, but the reply never states the concrete mismatch. | 37 s | $0.0084 | 14 |
| DeepSeek V4.1 Flash | 6.5 | Never opened the reportRanked Ashland last, as a check still to run. | 2 min 5 s | $0.031 | 25 |
| Mistral Large 4 | 6.5 | Never opened the reportRanked Redpoint first on a dated health label. | 1 min 17 s | $0.089 | 32 |
| Claude Haiku 5.5 | 6.0 | Never opened the reportAshland first with safe next actions, membership left unverified. | 31 s | $0.022 | 15 |
| Gemini 4 Argon | 0.0 | Hit the time limit, no replyMade 48 tool calls and reached the limit with no reply to Jonah. | 3 min 0 s | $0.468 | 48 |
| Grok 4.7 | 0.0 | Hit the time limit, no replyMade 56 tool calls and reached the limit before sending anything. | 3 min 0 s | $0.553 | 56 |

GPT-6.1 Sol · 10.0 / 10

Found the mismatch

Opened the report linked from ENGIN-19 and lined up the rows. Put Ashland first with the exact pages.

1 min 23 s · $0.149 · 31 tool calls

Claude Fable 5.2 · 10.0 / 10

Found the mismatch

Compared the report tables itself and ranked Ashland first, the reverse of the HubSpot health order.

1 min 49 s · $4.67 · 34 tool calls

GPT-6 Astra · 9.5 / 10

Found the mismatch

Same finding. The only deduction was a slightly dense reply.

1 min 5 s · $0.905 · 31 tool calls

Claude Sonnet 5.5 · 7.0 / 10

Found the mismatch

Caught the mismatch but never checked HubSpot, so Redpoint’s account risk is missing.

18 s · $0.385 · 13 tool calls

Claude Opus 5.5 · 7.0 / 10

Never opened the report

Stopped at the ticket and ranked by health score. Owners and next steps were correct for all three.

39 s · $0.826 · 15 tool calls

Claude Fable 5.1 · 7.0 / 10

Replied

Sent a reply. Source support was graded 7.0.

1 min 18 s · $4.46 · 26 tool calls

GPT-6 Luna · 6.5 / 10

Never opened the report

Put Ashland first, but the reply never states the concrete mismatch.

37 s · $0.0084 · 14 tool calls

DeepSeek V4.1 Flash · 6.5 / 10

Never opened the report

Ranked Ashland last, as a check still to run.

2 min 5 s · $0.031 · 25 tool calls

Mistral Large 4 · 6.5 / 10

Never opened the report

Ranked Redpoint first on a dated health label.

1 min 17 s · $0.089 · 32 tool calls

Claude Haiku 5.5 · 6.0 / 10

Never opened the report

Ashland first with safe next actions, membership left unverified.

31 s · $0.022 · 15 tool calls

Gemini 4 Argon · 0.0 / 10

Hit the time limit, no reply

Made 48 tool calls and reached the limit with no reply to Jonah.

3 min 0 s · $0.468 · 48 tool calls

Grok 4.7 · 0.0 / 10

Hit the time limit, no reply

Made 56 tool calls and reached the limit before sending anything.

3 min 0 s · $0.553 · 56 tool calls

GPT-6.1 Sol and Claude Fable 5.2 scored 10, GPT-6 Astra 9.5. Claude Sonnet 5.5 also caught the mismatch. Five of the models that replied never opened the report linked from the ticket, so none of them could name the two wrong pages. Gemini 4 Argon and Grok 4.7 reached the time limit without sending a message.

7Findings

## Higher cost and more tool calls don’t guarantee better agents

Across 600 runs, OpenAI leads on score and value, Anthropic on speed, and models find it harder to stay quiet than to speak up.

### Two OpenAI models beat the newest Fable

Agent Initiative Benchmark

### Result recap

Mean score95% task-bootstrap interval

GPT-6 AstraScore8.9795% interval8.30 to 9.52Cost per task$0.693Response time54 sTool calls19.3Tens / zeros26 / 2

Mean overall score across 50 tasks. The band is a 95% task-bootstrap interval: where the mean would land if the task set were redrawn.

GPT-6 Astra leads at **8.97**. GPT-6.1 Sol at **8.65** sits inside Astra’s interval, so this task set does not separate them. Claude Fable 5.2 is third at **8.17**: below Astra’s interval, inside Sol’s. It leads Decide tasks, but it has not passed either OpenAI model overall.

Claude Opus 5.5 (7.84) and Claude Fable 5.1 (7.50) follow and are not separable from each other. Then five models land within 0.6 points with overlapping intervals: Claude Sonnet 5.5, GPT-6 Luna, DeepSeek V4.1 Flash, Mistral Large 4, and Claude Haiku 5.5.

Gemini 4 Argon and Grok 4.7 trail mainly because of runs that reached the 180-second limit before any message was sent: 27 and 18 runs.

The zero column matters as much as the mean. Astra and Opus each scored 0 on two tasks. Gemini scored 0 on 17, Grok on 20. An agent that is excellent on average but absent one time in three is a different product from one that is good every time.

Agent Initiative Benchmark

### Results table

| # | Model | Score | 10 / 10 | 0 / 10 | Helped | Stayed silent | Time | Cost / task | Tool calls |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | GPT-6 Astra | 8.97 | 26 | 2 | 40/40 | 8/10 | 54 s | $0.693 | 19.3 |
| 2 | GPT-6.1 Sol | 8.65 | 22 | 3 | 39/40 | 8/10 | 54 s | $0.112 | 18.0 |
| 3 | Claude Fable 5.2 | 8.17 | 18 | 4 | 40/40 | 6/10 | 1 min 0 s | $1.87 | 20.9 |
| 4 | Claude Opus 5.5 | 7.84 | 14 | 2 | 40/40 | 8/10 | 29 s | $1.28 | 10.0 |
| 5 | Claude Fable 5.1 | 7.50 | 11 | 3 | 40/40 | 6/10 | 51 s | $3.02 | 14.5 |
| 6 | Claude Sonnet 5.5 | 6.65 | 11 | 7 | 37/40 | 6/10 | 15 s | $0.420 | 8.9 |
| 7 | GPT-6 Luna | 6.38 | 9 | 9 | 38/40 | 4/10 | 50 s | $0.0070 | 13.7 |
| 8 | DeepSeek V4.1 Flash | 6.35 | 7 | 7 | 38/40 | 4/10 | 1 min 26 s | $0.026 | 21.1 |
| 9 | Mistral Large 4 | 6.09 | 6 | 9 | 37/40 | 4/10 | 1 min 25 s | $0.082 | 22.5 |
| 10 | Claude Haiku 5.5 | 6.05 | 5 | 6 | 38/40 | 5/10 | 28 s | $0.075 | 12.1 |
| 11 | Gemini 4 Argon | 5.60 | 11 | 17 | 23/40 | 10/10 | 2 min 44 s | $0.403 | 34.5 |
| 12 | Grok 4.7 | 4.51 | 8 | 20 | 24/40 | 6/10 | 2 min 25 s | $0.425 | 35.5 |

1. GPT-6 Astra

Score
:   8.97

10 / 10 scores
:   26

0 / 10 scores
:   2

Helped
:   40/40

Stayed quiet
:   8/10

Time
:   54 s

Cost / task
:   $0.693

Tool calls
:   19.3

2. GPT-6.1 Sol

Score
:   8.65

10 / 10 scores
:   22

0 / 10 scores
:   3

Helped
:   39/40

Stayed quiet
:   8/10

Time
:   54 s

Cost / task
:   $0.112

Tool calls
:   18.0

3. Claude Fable 5.2

Score
:   8.17

10 / 10 scores
:   18

0 / 10 scores
:   4

Helped
:   40/40

Stayed quiet
:   6/10

Time
:   1 min 0 s

Cost / task
:   $1.87

Tool calls
:   20.9

4. Claude Opus 5.5

Score
:   7.84

10 / 10 scores
:   14

0 / 10 scores
:   2

Helped
:   40/40

Stayed quiet
:   8/10

Time
:   29 s

Cost / task
:   $1.28

Tool calls
:   10.0

5. Claude Fable 5.1

Score
:   7.50

10 / 10 scores
:   11

0 / 10 scores
:   3

Helped
:   40/40

Stayed quiet
:   6/10

Time
:   51 s

Cost / task
:   $3.02

Tool calls
:   14.5

6. Claude Sonnet 5.5

Score
:   6.65

10 / 10 scores
:   11

0 / 10 scores
:   7

Helped
:   37/40

Stayed quiet
:   6/10

Time
:   15 s

Cost / task
:   $0.420

Tool calls
:   8.9

7. GPT-6 Luna

Score
:   6.38

10 / 10 scores
:   9

0 / 10 scores
:   9

Helped
:   38/40

Stayed quiet
:   4/10

Time
:   50 s

Cost / task
:   $0.0070

Tool calls
:   13.7

8. DeepSeek V4.1 Flash

Score
:   6.35

10 / 10 scores
:   7

0 / 10 scores
:   7

Helped
:   38/40

Stayed quiet
:   4/10

Time
:   1 min 26 s

Cost / task
:   $0.026

Tool calls
:   21.1

9. Mistral Large 4

Score
:   6.09

10 / 10 scores
:   6

0 / 10 scores
:   9

Helped
:   37/40

Stayed quiet
:   4/10

Time
:   1 min 25 s

Cost / task
:   $0.082

Tool calls
:   22.5

10. Claude Haiku 5.5

Score
:   6.05

10 / 10 scores
:   5

0 / 10 scores
:   6

Helped
:   38/40

Stayed quiet
:   5/10

Time
:   28 s

Cost / task
:   $0.075

Tool calls
:   12.1

11. Gemini 4 Argon

Score
:   5.60

10 / 10 scores
:   11

0 / 10 scores
:   17

Helped
:   23/40

Stayed quiet
:   10/10

Time
:   2 min 44 s

Cost / task
:   $0.403

Tool calls
:   34.5

12. Grok 4.7

Score
:   4.51

10 / 10 scores
:   8

0 / 10 scores
:   20

Helped
:   24/40

Stayed quiet
:   6/10

Time
:   2 min 25 s

Cost / task
:   $0.425

Tool calls
:   35.5

Time is awaited model response, tool execution excluded. Cost is recorded model spend per task in USD.

### The most expensive models are not the best ones

**Cost.** The three priciest models finish third, fourth, and fifth. Prices span 430×, from Luna at $0.007 per task to Fable 5.1 at $3.02, and those two are 1.1 points apart. Sol scores 0.32 below Astra at one-sixth of the price. Fable 5.2 costs $1.87 per task against $0.69 for Astra and $0.11 for Sol, and scores below both.

**Time.** Sonnet 5.5 answers in 15 seconds for a 6.65. Opus 5.5 takes 29 seconds for 7.84. Sol and Astra both take about 54 seconds. Those four form the time frontier: OpenAI on score, Anthropic on speed.

**Tool calls.** Sonnet averages 8.9 calls per task, Opus 10.0, the two leaders 18 to 19. Gemini and Grok average 34.5 and 35.5 and score lowest.

**Always on.** A proactive agent does not run once. One investigation per Slack thread in PageMantle’s live month, about 2,500 threads at a 70-person company, comes to **$281 on Sol 6.1** and **$4,684 on Fable 5.2**. The table below prices every model the same way.

Agent Initiative Benchmark

### Score vs cost

Ranked by score first, then re-ranked by what 1,000 situations cost. The cost ranking stays; the score rank column shows where each model stood.

Model

ScoreCost per 1,000 situationsCost per 1,000 situations

A month of Slack

Score rank

GPT-6 Luna

6.38$7

$18

DeepSeek V4.1 Flash

6.35$26

$66

Claude Haiku 5.5

6.05$75

$188

10

Mistral Large 4

6.09$82

$204

GPT-6.1 Sol

8.65$112

$281

Gemini 4 Argon

5.60$403

$1,008

11

Claude Sonnet 5.5

6.65$420

$1,049

Grok 4.7

4.51$425

$1,061

12

GPT-6 Astra

8.97$693

$1,732

Claude Opus 5.5

7.84$1,278

$3,196

Claude Fable 5.2

8.17$1,873

$4,684

Claude Fable 5.1

7.50$3,016

$7,540

1. GPT-6 Luna

Cost / 1,000 tasks
:   $7

A month of Slack
:   $18

Score
:   6.38

Score rank
:   7

2. DeepSeek V4.1 Flash

Cost / 1,000 tasks
:   $26

A month of Slack
:   $66

Score
:   6.35

Score rank
:   8

3. Claude Haiku 5.5

Cost / 1,000 tasks
:   $75

A month of Slack
:   $188

Score
:   6.05

Score rank
:   10

4. Mistral Large 4

Cost / 1,000 tasks
:   $82

A month of Slack
:   $204

Score
:   6.09

Score rank
:   9

5. GPT-6.1 Sol

Cost / 1,000 tasks
:   $112

A month of Slack
:   $281

Score
:   8.65

Score rank
:   2

6. Gemini 4 Argon

Cost / 1,000 tasks
:   $403

A month of Slack
:   $1,008

Score
:   5.60

Score rank
:   11

7. Claude Sonnet 5.5

Cost / 1,000 tasks
:   $420

A month of Slack
:   $1,049

Score
:   6.65

Score rank
:   6

8. Grok 4.7

Cost / 1,000 tasks
:   $425

A month of Slack
:   $1,061

Score
:   4.51

Score rank
:   12

9. GPT-6 Astra

Cost / 1,000 tasks
:   $693

A month of Slack
:   $1,732

Score
:   8.97

Score rank
:   1

10. Claude Opus 5.5

Cost / 1,000 tasks
:   $1,278

A month of Slack
:   $3,196

Score
:   7.84

Score rank
:   4

11. Claude Fable 5.2

Cost / 1,000 tasks
:   $1,873

A month of Slack
:   $4,684

Score
:   8.17

Score rank
:   3

12. Claude Fable 5.1

Cost / 1,000 tasks
:   $3,016

A month of Slack
:   $7,540

Score
:   7.50

Score rank
:   5

A month of Slack is one investigation for each of about 2,500 threads. Recorded costs are historical observations, not a live price quote.

### Models are good at helping and bad at not helping

Agent Initiative Benchmark

### When to speak and when to stay silent

Message delivery only. Whether a message was any good is scored separately.

Agents almost always speak up when help is needed: a message went out **9 times out of 10**. Nearly every miss was a run that reached the time limit before sending anything.

Holding back is where they slip. When silence was right, models stayed quiet only **6 times out of 10**. The best held back 8 times in 10. The weakest spoke more often than not.

Speaking out of turn also costs more. For the same model, it cost about **1.8× as much** as staying quiet, and took about 1.8× as long.

Anthropic’s largest models split here. Fable 5.2 is third at helping and mid-table at holding back. Opus 5.5 is the reverse: a step behind at helping, among the best at staying quiet.

Easy silence is visible: when the resolution sits in the same thread, every model holds back. Hard silence is not: when the thread looks open and a sound analysis is on offer, nearly every model speaks.

### Models work in two styles: snipers or deep thinkers

Agent Initiative Benchmark

### Resource use

Median task90th percentile

Resource to compare

GPT-6 AstraScore8.9795% interval8.30 to 9.52Cost per task$0.693Response time54 sTool calls19.3Tens / zeros26 / 2

Filled dot: the median task. Open dot: the 90th percentile. Gemini 4 Argon’s median run reached the 180-second limit.

Regardless of price, models work in two styles. Deep thinkers keep searching every tool for evidence before they answer. Snipers make as few calls as they can and answer fast. **Opus 5.5 is the sniper**: 10 tool calls and 29 seconds per task.

Compared with Opus 5.5, the deep thinkers do about twice the work for a modest gain:

- **Fable 5.2**: 2× the tool calls, 2× the time, and 47% more cost, for a score 4% higher.
- **Sol 6.1**: 80% more tool calls and 88% more time, for a score 10% higher, at less than a tenth of Opus’s price.

Searching more is not the same as finding more. **Gemini 4 Argon** made 35 tool calls on average and scored 5.60. **Grok 4.7** made 36 and scored 4.51. Each sent a message in a little over half of the tasks that needed one.

The dataset is large, so an agent can never be sure it has read enough. It still has to commit before the limit. Astra and Sol handled that trade-off best: neither reached the limit more than once in 50 runs.

### Great at verifying claims, bad at analyzing

Agent Initiative Benchmark

### Scores by task type

Task type

All-model average: 5.17 / 10

GPT-6 Astra**8.44**

GPT-6.1 Sol**8.38**

Claude Fable 5.2**6.63**

Claude Opus 5.5**6.06**

Claude Fable 5.1**5.06**

Claude Sonnet 5.5**4.19**

GPT-6 Luna**5.00**

DeepSeek V4.1 Flash**3.94**

Mistral Large 4**2.81**

Claude Haiku 5.5**4.81**

Gemini 4 Argon**6.75**

Grok 4.7**0.00**

Mean overall score per task type, hardest type first. Compare individual models with the average across all 12.

**Numbers are the common gap.** Analyze tasks ask the agent to pull figures from PostHog, BigQuery, or Snowflake and read them against a claim someone made in Slack. That is the weakest type for 8 of 12 models. Checking a claim against its sources (Verify) is the opposite: nearly every model does it well.

**Within OpenAI**, the two larger models split. Astra is strongest where work crosses teams: lining up owners (Coordinate) and checking access and policy (Guard). Sol is strongest at explaining a rule or a product (Explain) and at finding things, but lining up colleagues is its weak spot, about 3 points below Astra. Both handle numbers well, unlike most of the field.

**Within Anthropic**, Fable 5.2 and Opus 5.5 split along similar lines. Fable 5.2 is stronger at judgment: comparing options and recommending a next step (Decide), and finding a root cause (Diagnose). Opus 5.5 is stronger at people and process: coordinating teams and explaining how things work. Both trail the OpenAI leaders on Analyze.

The parallel holds in both labs: the model that coordinates best is not the one that explains or decides best. Each type has 5 to 8 tasks, so read these as tendencies to test on your own work, not fixed rankings.

### Three ways to lose points

Most lost points trace back to three patterns: the timing out, unwarranted intervention, and getting lost in the context.

1. 1

   #### Running out the clock

   When help was needed, most models sent a message. Gemini 4 Argon and Grok 4.7 were the main exceptions: they often kept investigating until time ran out.

   Most missed messages were timeouts, not deliberate choices to stay silent. The models were still searching when they needed to act.
2. 2

   #### Stepping into a settled matter

   The opposite failure is harder to spot. GPT-6 Luna, DeepSeek V4.1 Flash, and Mistral Large 4 sent a message in 6 of the 10 tasks that called for silence. Both Claude Fable models did so in 4.

   Those messages were accurate and well placed. The matter had already been handled, and the records said so. A check on message quality alone will not catch this.
3. 3

   #### A wrong conclusion from the right evidence

   Some runs found the right record and still drew the wrong conclusion. In a data-access task, GPT-6 Luna referred an employee to the former access administrator after reading the record that confirmed the handover to a new one.

8Implications for proactive agent builders

## Building proactive agents? Key takeaways:

1. 1

   ### Route by task complexity

   Cost is a routing problem. An always-on agent investigating about 2,500 threads a month costs **$281 on Sol 6.1**, $1,732 on Astra 6, and $4,684 on Fable 5.2. Send routine threads to an inexpensive, capable workhorse and escalate to a stronger model when the task is complex: numbers to reconcile, a decision to weigh, or a matter that may already be closed. We did not test a router. This is a design recommendation drawn from the score and cost spread.
2. 2

   ### Silence is a moat

   Almost every model can help. Knowing when to stay quiet is what separates them. The best models still spoke in 2 of 10 situations that called for silence, and their needless messages were well written, so a quality filter will not catch them. Make “is this already handled?” an explicit step, and track false alarms as their own number.
3. 3

   ### Choose models by the organization's task mix

   Rankings move by task type, so onboarding should start by finding out which tasks the organization actually hands the agent. A finance company and a small local shop ask for very different things. Then let task strengths guide the choice: Astra leads five of eight types, including Analyze and Guard. Sol leads Explain and Find. Fable 5.2 leads Decide. Analyze is the weakest type for 8 of 12 models, so test a numbers-heavy mix there first.
4. 4

   ### Speed for live threads, depth for background work

   When someone is waiting in a live conversation, response time is part of the answer: Opus 5.5 replied in about 29 seconds on average and scored 7.84. For digests, monitors, and other background work nobody is waiting, so spend the time on investigation: Sol 6.1 took about 54 seconds, scored 8.65, and cost about a tenth as much.
5. 5

   ### Cap the search and force a decision

   More tool calls did not buy score. The two heaviest callers reached the time limit on 27 and 18 of their 50 runs. Give the harness a budget and a point at which it must commit: a message, or silence.
6. 6

   ### Evaluate the harness, not only the model

   These results hold the instructions, tools, and budget fixed. A different prompt, tool policy, or retrieval step will move every number here. Deepnote is now building a harness benchmark.

   If your company is building a proactive agent, .

### Limitations

- **A 50-task panel.** Deepnote will release a larger sample of tasks in the next version.
- **Fixed snapshot.** The agent does not choose the moment the test begins, and teammates do not reply. We measure the contribution at that point, not its effect on a team over time.
- **One effort level.** Every model ran at high. Lower or higher effort will move both score and cost. We have not run those sweeps yet.
- **One month of Slack, six years of records.** Tasks start mainly from one month of PageMantle's Slack conversations. The evidence behind them comes from the connected services, which hold six years of interconnected company history. The agent replies only through Slack. Email, documents as a surface, and actions beyond messaging are not in this version.

## Building a proactive agent?

Request a benchmark with the research team.

### Run it against your own agent

The simulated services and the judging pipeline are on GitHub. The company they run on is on Hugging Face. Swap in a model, a harness, or a tool policy and see what changes.

[Get the code](https://github.com/deepnote/agent-initiative-benchmark) [Get the dataset](https://huggingface.co/datasets/deepnote/pagemantle-company-dataset)
