Hacker News Reader

No free models are even in the same class as Claude when it comes to code. Investors are clearly confident that moat won't be crossed soon.

According to Aider's leaderboards, R1 is not that far behind Sonnet (without thinking): 56.9% vs 60.4%

Sonnet with thinking is at 64.9% (the top), but it's pricey - about double the cost without thinking.

In my experience DeepSeek is far better than Claude for coding.