I ran Fable 5.1 on the crpg-bench of "Jin Yong Heroes" and it didn't feel very amazing.
I ran Fable 5.1 on the crpg-bench of "Jin Yong Heroes" and it didn't feel very amazing. There is no significant difference in IQ level between playing games and fable5/opus5/sonnet5. Jin Yong Heroes Bench is what I use to test large models in long-horizon
I gave Fable 5.1, Fable 5, Claude Opus 5 and Codex the exact same prompt
Build an interactive japanese koi pond. The results surprised me. Fable 5.1 was the clear winner. The water, ripples and overall motion looked noticeably more realistic, but it did cost more than Fable 5:
1,001 very messy rows in a CSV file.
A single prompt with clear rules for how to handle issues. Two AI agents: claudeai with Opus 5 vs. opencode with GPT-5.6 Luna (via OpenRouter). Neither followed the instructions perfectly. One got a perfect score.
Opus 5 vs Fable 5 vs Fable 5.1
Compare with same prompt ・Opus outputs “Claude-like = AI-like” ・Fable 5 also feels like a document ・Fable 5.1 is established as an illustrated animation Especially the middle to second half👀
You have reached the end of the archive
All of Claude Opus 5