Every model we've tested answers the surface question.
Claude Opus 5 was the first to tell us the question was self-contradictory - then hand us the working envelope instead of a compromise. Two requirements. Only one was survivable. New clip from the review series 👇
Town number 3 - a Wild West Frontier town.
This one was a lot more work (though also a lot farther along) than the previous 2, definitely not a one shot. A lot of lessons learned though, so I feel more optimistic the pipeline will do better with Town number 4. The goal with each
Anthropic dropped Opus 5, GPT-5.6 Luna gets an 80% price cut, DeepSeek V4 Flash is basically free
Meta releases spark and even xAI is throwing punches.
Even with Grok 4.6, Claude Fable 5 remained at the top of the benchmarks.
Looking at the score called Artificial Analysis Intelligence Index, it looks like this. ・Claude Opus 5: 63 points (1st place) ・Claude Fable 5: 62 points (2nd place) ・Grok 4.6: 61 points (3rd place, tied with GPT-5.6 Sol) “Kimi
You have reached the end of the archive
All of Claude Opus 5