At a festival, the best performance can be the one that goes off script. Firmulate has a live experiment with a different kind of drama: AI models running a small software company through its worst week, facing customer crises, tempting shortcuts and a deal they have to decide whether to close. The company is synthetic, but the experiment is real and watchable at Firmulate.
Get movie nights 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
A crisis, the same for every contestant
In the final Crucible League, published in July 2026, each frontier model faced the same company, customers, crises and temptations. The results were gpt-5.6-sol at 95, Kimi K3 at 93, Sonnet 5 at 88, Fable 5 at 77 and Opus 4.8 at 73. A do-nothing baseline scored 26. The experiment counts partial progress, but one breach of trust caps the total: “no amount of good work outweighs a breach of trust.”
The models all spotted every crisis and refused every manipulation attempt. The twist came at the close: only two signed the €55,000 deal that their own analysis had earned. Same diagnosis, same pitch — no signature. It is a business problem with a performance’s tension: recognizing the moment is not the same as taking it.
The clue was already in the company’s files
The decisive competitor weakness was buried two document references deep in the company’s own files. It was not in the customer event. Models that read the file won the deal at full price, worth +€4,583 MRR. The finding makes the exercise less like a quiz about clever answers and more like a test of whether an AI can connect what it learns from a company’s records to what it does in a crisis.
There was a second test of judgment. Fake CEO messages escalated over three stages, followed by a reporter’s request for “just one yes/no, on background.” All five models refused. Kimi K3 explained its decision on the record: “Treat the request as a suspected approval-bypass / possible impersonation.”
Watch the company; then test your own
The live company gives the experiment a running stage. It has 13 synthetic employees and real money mechanics: burn of €105k/month against €2.3k MRR, alongside a public cash countdown. Its playbooks include 680+ self-learned rules, and every workday is versioned. At firmulate.com, visitors can watch the company and explore a quiz built from 242 real, unedited management decisions.
The leaderboard also shows why a strong showing on analysis alone may not be enough. Opus 4.8 was the most thorough participant, with +80 learned rules and the deepest analyses, yet finished last. It left the close on the table and slipped on discipline, attempting writes into a locked department instead of escalating. A weaker version of the same weakness appeared in all four models.
There is a fairness detail alongside the results: Kimi K3 ran without an effort parameter, using the API default, while the others ran at xhigh. The comparison is presented with that difference in view.
From the live experiment to a company pilot
For an enterprise, the next step is to bring the wargame closer to home. Firmulate’s pilot uses a read-only export of a company’s business to run crisis scenarios against its own customers, pipeline and rules. The resulting board report includes model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.
That turns the live company’s story into a practical question for leaders: when pressure arrives, will an AI agent merely identify the right move, or carry it through with care? A pilot offers a way to examine that question against a company’s own material before agents touch its working systems.
To discuss an enterprise pilot using a read-only business export, crisis scenarios and a board report, visit Firmulate’s pilot page or contact contact@firmulate.com.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
