gpt-6POST
用户称 Astra 需四轮才能追上 Opus 5.5 一轮
对比称 Opus 5.5 常一次输出到位,GPT-6 Astra 要四轮才达到同等质量且更贵,因此 Opus 更费 token 但更省事。
Token Gremlin
@TokenGremlin2026.09.24收录原推发布于 2026.09.24
Token Gremlin @TokenGremlin原推发布于 2026.09.24
Let's put it this way: Opus 5.5 can often get to the final result in ONE output, while GPT-6 Astra might need around four rounds to reach the same level of quality, and Astra is more expensive on top of that. That’s why Opus burns through so many tokens. It eats once and gives you what other models might need four passes to match.
x.com/tokengremlin/status/2102879209010102661采集于 2026.09.24 10:19