I Just tried the New GPT 6 SOL on a Crazy Benchmark !

Hey Guys, GPT-6 Sol is here — but how good is it compared to GPT-6 Astra? In this video, We will see what GPT 6 Sol brings to the Table . What are its benchmark against models like claude and Other open source Fronteir models. I will also test GPT-6 Sol and GPT-6 Astra on two very different real-world tasks to see how they actually perform beyond benchmark numbers. For the first test, I gave both models a much more complex task: creating a motion graphics animated project directly from a script. For the second test, I gave them a more generic coding task: creating a GTA Vice City-inspired browser game with Minecraft-style voxel graphics. Instead of only looking at the final output, I compare both models based on: • Quality of the final result • Coding and execution ability • How well they understand the task • Token usage • Time taken to complete the task • Stability and consistency • How much work they can complete within the available usage limits I also compare the final projects side-by-side so you can see the actual difference between GPT-6 Sol and GPT-6 Astra. The interesting question isn't only which model is more intelligent. For agentic coding, efficiency, stability, token usage, and the ability to keep working for a long time can be just as important. So, is GPT-6 Sol a meaningful upgrade? And how close does it get to GPT-6 Astra in real-world work? Watch till the end to see the final results. If you enjoy AI model comparisons, coding experiments, AI agents, vibe coding, and testing new AI tools on real projects, consider subscribing. Models tested: GPT-6 Sol and GPT-6 Astra Tests: Motion graphics generation + browser game development #GPT6 #OpenAI #AICoding #GPT6Sol #ArtificialIntelligence #VibeCoding #AIAgents #CodingWithAI