It must be some kind of psy-op by Anthropic to claim that Fable is anywhere near as good as Astra
- 发布
- 09/05 08:44
- 收录
- 09/05 15:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 同步延迟
讨论趋势
百分比基于采集到的讨论信号，不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。
一位Reddit用户，同时订阅了ChatGPT Pro和Claude Max，声称Anthropic的Fable 5.1模型与ChatGPT的旗舰模型Astra相比，性能显著逊色，尽管两者的价格大致相当。该用户认为Fable 5.1是一个“过时模型，基于已死的架构框架”，并指出Astra提供了更好的价值，即使其价格昂贵且有每周使用限制，但其卓越的输出质量和直观的思维能力使其物有所值，而Fable则无法与之相提并论。
I have a ChatGPT Pro subscription and a Claude Max subscription, and use both extensively for work. To claim that any model offered by Anthropic is even close in intuitive thinking or design sensibility to Astra is a joke to me.
The most comparable Antropic model to Astra, ChatGPT's flagship, is Fable 5.1. They have roughly equivalent price. But in my experience, their output quality is dramatically different.
Astra has a sort of deep, intuitive understanding about the problem you give it that Fable simply lacks. When I compare how the two work, Fable feels like it gives the smartest answer you could possibly get after asking 100 moderately smart people to deliberate together in a room for a few hours before answering. Astra just feels... smart. Like a single, extremely smart person giving an intuitive answer that fully understands the question and work statement.
Where Fable feels like it just spends a lot of time talking to itself in order to work, Astra just...works. And does better than Fable. Its answers are easier to understand, not held back by the limitations of the language it's using to speak with.
To me, to say Claude Fable 5.1 is anywhere close to Astra in any capacity, and not just an outdated model on a dead architectural framework, is completely unfathomable to me. You pay the same price for double the weekly usage, so you more than get your money's worth. Sure it's expensive, and you can run through your weekly limits in hours, but you can't argue that it just works. I can't say the same about Fable.