Ask HN: What was the last task where only a frontier model could do it?

Hacker News is crowdsourcing a reality check on AI model capabilities. A developer posted a question to the community: when was the last time a frontier model—like Claude Opus or GPT—actually solved something that open-source alternatives couldn't? They're pushing back against claims that models six months behind the cutting edge handle ninety-five percent of real-world tasks. The call is for concrete examples from the past month—specific work where DeepSeek or Qwen failed, but frontier models succeeded. It's a timely question as open-source models improve rapidly and the capability gap continues to narrow.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton