A 30% model claim can make Opus look like the loser before the test starts.
But Anthropic compared Sonnet 5.5 with Sonnet 5, not Opus 5.5. Token prices answer a different question. Would you choose a model from a launch chart or test your own task first?
Damn ❗️ How did you create that fire effect? I was shocked. Opus 5.5 became a teacher.
Sonnet 5.5 work, Step 5 preview provides materials, Happy Shrimp generates music
Opus 5.5 made this with 5 prompts + 2 /goal + 20 hours + 2.94B tokens + $60 seeddance + $20 elevenlabs.
(sound on 🔉) full prompts:
I went back and forth between Chat and Code for Opus 5.5
I went back and forth between Chat and Code for Opus 5.5, asking for reviews and implementation, and the browser game I was working on today ended up looking like this. There is almost no rework, and the quality is better than I expected. I wonder how far I can go into it. On the Pro plan, I used up the entire 5-hour time slot at a time.
You have reached the end of the archive
All of Claude Opus 5