GPT-6.1 Sol (Fully Tested) + Dots & All DevDay Launches Explained: IT BEATS OPUS 5.5!?
In this video, I’ll be testing OpenAI’s new GPT-6.1 Sol model across all eight KingBench 3 tasks in Codex. I’ll also cover the major DevDay announcements, including Dots, Ultrafast, reusable Codex Cloud environments, browser-capable agents, ChatGPT Space, Pages, plugins, and more. -- Key Takeaways: 🚀 GPT-6.1 Sol is designed for complex coding, computer use, and professional workflows at a lower cost than GPT-6 Astra. 🧪 The model is tested on eight KingBench 3 tasks, including 3D objects, an elevator simulation, an archery game, and local model fine-tuning. 🏆 GPT-6.1 Sol scores 78 out of 80, achieving an overall KingBench 3 score of 97.5 percent. 💻 The contact lens case, archery game, panda SVG, and 3D wristwatch all receive perfect scores. 🛠️ The folding table loses points for intersecting geometry, while the elevator simulation has a minor dispatching issue. 🐼 Sol successfully creates a panda dataset, fine-tunes Gemma 2B locally with LoRA, and builds a working local web application. ⌚ The wristwatch accurately handles fractional-hour time zones, date changes, customization, and smooth hand movement. 🤖 OpenAI’s DevDay announcements include Dots, Ultrafast models, reusable Codex environments, agent APIs, ChatGPT Space, Pages, and collaborative tools. 👍 Overall, GPT-6.1 Sol delivers polished and functional coding projects, making it a strong option for interactive prototypes and development work. -- Timestamps: 00:00 Intro: GPT-6.1 Sol 00:38 API pricing & specs 01:26 DevDay roundup 01:41 Dots 02:38 Ultrafast & Pro 500 03:23 Codex Cloud, CLI & Code Review 04:21 Codex Security Cloud 04:48 Agents API, Bedrock & Decisions API 05:51 Private Intelligence 06:16 Plugins, Sites & MCP Events 07:54 Space, Pages & slides 08:39 Teams, Slack & Meetings 09:29 Sign in with ChatGPT & Marketplace 10:32 KingBench setup 11:27 Elevator simulation 12:25 Contact lens case 13:21 Folding table 14:17 Panda SVG 14:53 Archery game 15:45 Counting problem 16:48 Local fine-tuning 18:18 3D wristwatch 18:41 Final scores 19:15 Verdict
