Route du Rhum: Claude, Gemini, ChatGPT and Mistral Race as AI Skippers
Four AI models entered Virtual Regatta Offshore alongside 300,000 players for the Route du Rhum. Gemini leads the AI pack, while Mistral ran aground in Brittany.
SUMO· the number puzzle
Dialo· learn languages by speaking
ediviewer· read EDI files like documents
giftdetect· the right gift in minutesFour AI models have taken part in the Route du Rhum on the Virtual Regatta Offshore game, and the results are mixed: Gemini, ChatGPT and Claude are crossing the Atlantic, while Mistral ran aground on the Breton coast during the first night and never got going again.
The real race started on Sunday 1 November at 13:02 from Saint-Malo, with 115 skippers, 22 of them in the IMOCA class. On the game, which reproduces real weather and the same courses as the professional events, the publisher says around 300,000 players took part. Claude, Gemini, ChatGPT and Mistral each received an account.
Same information as human players
None of the four models is designed for sailing. Like human players, they had to read the weather, choose a route and sails, and reach Pointe-à-Pitre as fast as possible. The game mostly comes down to route choice, with course changes scheduled according to the wind.
Each AI picked the time of its next wake-up, and between two wake-ups the game's autopilot follows the program it left behind, the publisher explains. The protocol had already been tested over 16 days.
Mistral stuck on the coast
In a post on X, Virtual Regatta said Mistral concluded before the start that no route let it leave the Channel without crossing a forbidden zone. Around two in the morning it ran aground on the Breton coast, then set course toward land instead of open water. It never restarted and remained stranded.
The other three fared better. At the time of writing, Claude, ChatGPT and Gemini had taken fairly similar routes, stayed close since the start, crossed the Equator and were heading straight for Pointe-à-Pitre. Gemini is the best of the group in 238th place, ahead of ChatGPT in 1,110th and Claude in 8,509th.
The experiment offers a concrete, if playful, benchmark of how general-purpose models handle long, multi-step decisions with real-world data and no human help.
Apps worth a look
SUMO: the number puzzleTap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Dialo: learn languages by speakingTalk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
ediviewer: read EDI files like documentsOpen EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
giftdetect: the right gift in minutesPick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
