Explore / Case study / StartLux vs Jev: One Snake Match

StartLux vs Jev: One Snake Match

Demo onlyWeb

A recorded 20×20 Snake comparison between local StartLux 27B and Jev 1.14.

by admin · listed 4 days ago

⚡ What System1 does here

According to the recording, each model chooses the snake’s next direction from game-state information. The game executes movement, places food, detects collisions and computes the score. The video shows a single run rather than a multi-seed evaluation.

About

A vertical recording from Decision Arena showing local StartLux 27B and Jev 1.14 playing separate Snake boards in parallel. The on-screen rules use a 20×20 board and award 100 points for each piece of food. The presentation labels the two sides and follows the match through the final result. In this recorded game, the result screen reports 700 points and 7 food for StartLux versus 100 points and 1 food for Jev. It lists survival times of 28 and 15 game seconds respectively: StartLux ends by hitting its own body, while Jev hits the boundary. These are the displayed outcomes of this one match. Evidence boundary: model identities and local execution are presented by the recording; this publication did not rerun the game or inspect a complete model-call trace. One result is not evidence of general model superiority, typical latency or performance across seeds and games. The original Chinese video and its on-screen result are retained.

StartLux vs Jev: One Snake Match ↗

🧪 Community test reports

0 reports · 0 passed on the real model
No one has tested this yet. Be the first.

💬 Discussions

New post →
No discussions yet.