https://peertube.watch/videos/watch/bdc45304-e978-4129-8262-d5516299828c Peace Call. Mr. Peace is an AI-made character; voice is AI-generated; script written with AI assistance. No affiliate links in this video. How this site is run: https://getpeacecall.com/verify/?utm_source=peertube&utm_medium=video&utm_campaign=underdog-doom In one DOOM benchmark (defend_the_center), a specialized 1.3-million-parameter model averaged 17.8 frags per game vs 0.6 for Nemotron-120B (our per-game arithmetic from the paper's totals; different numbers of games were played, see below). Agents got text (ASCII) frames + depth maps and the same 4 actions. Conditions: the small model was trained on ~2 hours of human play for this one scenario; it decides in 31 ms vs 8.9 s for Nemotron-120B; the reasoning LLMs played fewer games. In the paper's slower 'fair mode' the small model still led (10 frags; best LLM 5), but GPT-4o-mini survived nearly as long (avg 30 vs 33 steps) and had a higher max survival (55 vs 46). Source: Golchinfar, Vaziri, Marquardt, 'Playing DOOM with 1.3M Parameters: Specialized Small Models vs Large Language Models for Real-Time Game Control', arXiv:2604.07385v1 - PREPRINT (Apr 8, 2026), not yet peer reviewed. https://arxiv.org/abs/2604.07385 . The authors built the small model. Numbers from Tables 2-3. DOOM and model names are trademarks of their owners; not affiliated with id Software, NVIDIA or the authors. No game footage is used. Charts drawn by code from the paper's numbers. Mr. Peace is an AI-made character with an AI voice. The small model was evaluated over 10 episodes and Nemotron-120B over 5. Both were tested in one scenario using text/depth representations, with different input encodings. The small model was trained specifically for this task. The study was conducted by its creators and has not been peer-reviewed. Note: Under a "fair mode" setting (frame_skip=20, Table 3) that gives LLMs more time between decisions, GPT-4o-mini's survival time approaches the small model's. Full conditions in the paper: arXiv:2604.07385 Paper: https://arxiv.org/abs/2604.07385 (version v1, posted Apr 8, 2026). AI Disclosure: Mr. Peace is a fictional AI character using a synthetic voice. This video is AI-assisted and does not depict a real person speaking. Research Disclosure: Based on a research preprint that has not been peer-reviewed. Results apply to the reported benchmarks and experimental conditions, not overall model superiority. Trademark Disclosure: Product, model, and game names are used for identification and commentary. Trademarks belong to their respective owners. Peace Call is not affiliated with or endorsed by those companies. Tue, 03 Nov 2026 16:17:11 GMT https://validator.w3.org/feed/docs/rss2.html PeerTube - https://peertube.watch https://peertube.watch/lazy-static/avatars/3b284627-b75f-409e-ac2e-10f4c2946f43.png https://peertube.watch/videos/watch/bdc45304-e978-4129-8262-d5516299828c All rights reserved, unless otherwise specified in the terms specified at https://peertube.watch/about and potential licenses granted by each content's rightholder.