Kiana Ehsani(@ehsanik):55.7% -> 70.6% on OSWorld v2 (released just a few weeks ago) Genuinely would love to have a harder computer use evals. Happy to sponsor a computer use benchmark that Opus 5 gets below 50% on! If you have an eval that matches this criteria please reach out! Happy to chat.

Kiana Ehsani

@ehsanik

Making models smarter @ Anthropic, formerly CEO and Co-Founder @ Vercept (acquired by Anthropic), Climber on the weekends. Opinions are my own.

加入 March 2015

649 正在關注 11K 粉絲

Kiana Ehsani@ehsanik

2026.07.24 20:24

55.7% -> 70.6% on OSWorld v2 (released just a few weeks ago) Genuinely would love to have a harder computer use evals. Happy to sponsor a computer use benchmark that Opus 5 gets below 50% on! If you have an eval that matches this criteria please reach out! Happy to chat.

142

轉發到社區

熱門用戶