传闻：OpenAI 已完成 10T+ 参数模型 Bel，定位为后 GPT-6 基座
帖文称 OpenAI 已完成名为 Bel 的 10T+ 参数模型，可能作为 GPT-6 之后的新基座，并对比了 GPT-6 Astra 在 FrontierMath、ARC-AGI-3 等基准上的高分。
GPT-6 BEL MIGHT BE THE MODEL OPENAI DOESN’T WANT YOU TO KNOW ABOUT 💀 There’s a new rumor making the rounds… OpenAI has reportedly finished “Bel”, a model claimed to have 10T+ parameters. And the crazy part? It’s reportedly being positioned as the post-GPT-6 foundation 👀 Potentially even a foundation for an AGI-threshold model. But here’s where it gets interesting… There are NO public Bel benchmarks yet. Compare that with GPT-6 Astra, which already scores: 98% → FrontierMath Tier 4 99.9% → ARC-AGI-3 100% → ExploitBench So if Bel is actually the next major pretraining leap… those numbers could look very different. 💀 OpenAI hasn’t confirmed Bel. But if this rumor is real… GPT-6 might already be yesterday’s model. 👀