Press Space to continue
Claude's agent benchmarks still need significant improvement
Press Space to continue
People want Claude at home. The reality is that this is not it. Bring those long-horizon agent benchmarks up towards 98% in v3 please! > Headline: 98.2% retained in aggregate > Reality: 75.6% on swe-bench + terminal-bench. https://x.com/casper_hansen_/status/2101…
@PrismMLQuoted post unavailable hereOpen on X
open on x · 1 viewViews and clicks count once per visitor per day. Clicks are opens of this post on X.