北京时间8月8日凌晨，OpenAI发布了它们最新一代的GPT模型——GPT-5。 | | GPT-5 | Gemini 2.5 | Grok | Claude 4.1 | | --- | --- | --- | --- | --- | | | (high) | Pro | 4 | Opus | | AIME '25 (no tools) | 94.6% | 93.8% | 90.5% | 94.1% | | FrontierMath (with python tool | 26.3% | 27.1% | 24.0% | 25.8% | | only) | | | | | | GPQA diamond (no tools) | 85.7% | 86.1% | 83.2% | 85.9% | | HLE[1] (no tools) | 24.8% | 23.5% | 21.1% | 24.2% | | HMMT 2025 (no tools) | 93.3% | 92.9% | 89.7% | 93.0% | GPT-5以个位数优势领先竞争对手 这种合成数据的新应用，让前一代先进模型生成高质量数据，让后 ...

So I think synthetic data is actually really useful in some places, but I think people overestimate what it can do. I'll give a couple examples. So right now there are a bunch of models that have been trained really heavily on synthetic data, but like I mentioned earlier, it means that they're only good at very academic homework style benchmark style problems.They're actually terrible at real world use cases. So yeah, synthetic data, it's made models good at synthetic problems, not real ones. And we actuall ...