开发者吐槽 Astra 更省 token 但额度烧得更快

Rajath Gowda @buildwithrajath原推发布于 2026.09.06

在 X 上打开原推

🚨 GPT-6 Astra’s “efficiency” has me scratching my head. I keep hearing that Astra can get more work done with fewer tokens. Great. But as someone actually building with AI, the thing I notice is my usage disappearing faster than ever. 😭 That feels backwards. If the model is more efficient, I should be able to ship more, build more, and run more workflows before hitting my limit. Instead, we get: Better model. Fewer tokens. Same wallet. Faster usage burn. Maybe the economics make perfect sense for OpenAI. But as a SaaS founder paying for these models, I have a pretty simple question: Where do the efficiency gains show up for me? Because right now, I’m not feeling them. 💀