firmulate.com/benchmarks.html — live view
AIThis post was created with the assistance of artificial intelligence (AI).
Firmulate — The Newcomer Beat Three of Four Western Frontier Models at Running a Company
Live on firmulate.com.

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get furniture and decor delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

How Artificial Intelligence Is Changing Business Oversight

In today’s fast-paced business environment, decision-making tools are more critical than ever. For interior designers and furniture retailers, the question isn’t just about aesthetics, but about how AI can help manage crises, optimize workflows, and ensure trustworthiness. Recently, a public experiment revealed a newcomer AI model that outperformed several long-standing industry leaders in a rigorous management challenge — shedding light on the potential of emerging AI to revolutionize business operations.

The Live Experiment: Putting AI to the Test

Firmulate’s live experiment simulates a small, real-world software company navigating its most challenging week. The same crises, customer demands, and temptations were presented to multiple AI models, each running the company independently. The goal? To see which AI could best diagnose issues, resist manipulation, and close lucrative deals — all without human intervention.

Among the models tested, a newcomer called Kimi K3 from Moonshot scored a remarkable 93 out of 100 in the Crucible league, placing just behind the top scorer, gpt-5.6-sol, which scored 95. The other participants, Sonnet 5, Fable 5, and Opus 4.8, scored 88, 77, and 73 respectively. The results are especially striking considering all models faced identical scenarios and restrictions.

Key Findings: Trust and Attention to Detail Matter

The experiment’s core insight was that all models identified every crisis and refused manipulation attempts, demonstrating a high level of honesty and discipline. Yet, only two models managed to secure the €55,000 deal, which was earned through their own analysis and diagnosis. The winner, K3, found a buried piece of critical information—located two document references deep in the company’s files—that was essential for closing the deal at full price, worth an additional €4,583 in monthly recurring revenue.

This highlights a vital aspect: the ability to read and interpret supporting documentation is crucial for sound decision-making, often overlooked in less comprehensive AI assessments.

Resisting Manipulation and Maintaining Integrity

In scenarios involving social engineering, all models refused to be duped by fake CEO messages and a staged reporter trick. Kimi K3 explicitly treated these requests as potential impersonation or approval bypass attempts, reflecting a cautious and disciplined approach. Such resilience to manipulation is vital for AI systems operating in real business contexts where trust and security are paramount.

Beyond the Scores: The Real Business Impact

The experiment isn’t just about scores. It’s a window into how AI models could impact real companies, like those in interior design or furniture retail, where every decision can influence brand reputation and financial health. The live setup features 13 synthetic employees managing actual money mechanics, burning €105,000 monthly against a tiny €2,300 in monthly revenue, illustrating the high stakes involved.

For companies considering AI adoption, the key takeaway isn’t just whether the technology generates convincing chat responses. Instead, it must demonstrate consistent discipline, thorough analysis—including reading files properly—and honesty under pressure. The question for decision-makers is: which model will truly complete the work and sustain trust?

The Role of Fairness and Testing Standards

The experiment also highlights the importance of testing parameters. K3 ran without an effort parameter (the API default), while others ran at xhigh. Despite these differences, K3’s performance remains impressive, emphasizing that a thorough, disciplined approach can outweigh aggressive tuning.

Watch the Live Performance

Curious how these models operate in a real-time environment? The live experiment is accessible at firmulate.com/live. It showcases a real company managing crises daily, with decisions versioned and auditable—an invaluable tool for businesses contemplating AI integration in their operations.

Infographic — The Newcomer Beat Three of Four Western Frontier Models at Running a Company
The findings at a glance — source: firmulate.com.
Amazon

AI decision-making software for business

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Final Thoughts: Choosing the Right AI Model Matters

This experiment underscores a crucial point for any business: the most capable AI isn’t necessarily the one with the flashiest chat. It’s the one that completes tasks reliably, reads supporting documents thoroughly, and resists manipulation under pressure. For interior designers, furniture retailers, or any enterprise seeking trustworthy AI, the choice of model could mean the difference between a lost opportunity and a lucrative deal. The league is open, and testing your own AI before deploying it in critical workflows is now more essential than ever. Visit our benchmarks page to learn more, and consider running your own wargame at firmulate.com/pilot.html.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html

Powered by Thorsten Meyer AI


Amazon

business management AI tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Amazon

AI cybersecurity and fraud prevention software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Amazon

AI document analysis tools for business

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Teachers Hate This Free Learning Game – Find Out Why!

Uncover the surprising reasons why teachers dislike this popular free learning game and how it may impact students’ educational experiences.

Hollywood Stars: Typecast or Diverse Talents

Lights, camera, action! Explore Hollywood stars' journey from typecast roles to diverse talents, unraveling the dynamic world of entertainment.

Wohngeld

The German government plans to reform Wohngeld to increase support for low-income households, with new eligibility criteria and higher benefit caps.

Transform Lives: The Power of Empowering Others

Prepare to discover the profound impact of empowering others and unleashing their potential, creating a ripple effect of positivity.