GPT-7 Bel Huge Breakthrough, Claude Did It Too - on Same Day
In latest AI developments, OpenAI's secret internal model, most probably GPT-7 Bel, has disproved a 27-year-old belief in computer science, while Anthropic says Claude found a different way to do the exact same thing on the very same day. Meanwhile, a mystery OpenAI model in Arena could be GPT-6.1 Astra, Google's Gemini 4 Hydrogen has leaked alongside the new Genie 4, and DeepSeek V5, GLM 5.5 and Qwen 4 are getting close. First, we discuss what the Exponential Time Hypothesis is, in plain words, and what GPT-7 Bel actually proved about 3-SAT and NP problems. Then we look at its other two discoveries, about waves and the strong force, and how Claude reached the same ETH result by a different route. After that, we discuss whether GPT-6.1 Astra is about to come back, and what the scoreboard and the prediction markets say. Next, we look at Gemini 4 Hydrogen and what Genie 4 can do, and at the new models lining up in China. Finally, I share my honest opinion on these updates, and the caveats to keep in mind. This is part of my quest to find the best AI model in 2026. Subscribe to the channel, as I cover every update related to the top models, be it from the American or the Chinese AI labs. Timestamps: 0:00 - GPT-7 Bel and Claude Same Day Breakthrough 1:14 - What GPT-7 Bel Proved: ETH and NP Problems 4:56 - Waves, Atoms and Claude's ETH Proof 8:38 - Is GPT-6.1 Astra Coming Back? 10:48 - Gemini 4 Hydrogen Leak and Genie 4 12:36 - DeepSeek V5, GLM 5.5 and Qwen 4 14:39 - My Honest Opinion #GPT7 #GPT7Bel #OpenAI #Claude #Anthropic #ExponentialTimeHypothesis #GPT61Astra #Gemini4 #Genie4 #DeepSeekV5 #Qwen4 #AINews #ArtificialIntelligence #AIModels
