Positive users praise Cognition's FrontierCode leaderboard for tracking merge-worthy AI code over synthetic benchmarks while negative users call the charts hard to read and insult models like Sonnet 5.
Based on 29 visible X reactions from 67 accounts; directional sample.
Ask a question below.
Published answers will appear here.
The chart features a controversial backward horizontal axis for output tokens.
@cognition This is the kind of benchmark developers actually need. Real-world coding quality matters more than synthetic scores.
@cognition Com essa criação desse leaderboard é um passo excelente e muito necessário para o mercado de tecnologia.
@cognition Code you'd actually merge is a much better bar than most benchmarks use.
@cognition man who makes these stupid charts, hard to look at
@cognition Only thing I see here is how much of a dogwater Sonnet 5 is
@scaling01 Cursor Grok model is pretty underrated.
The FrontierCode leaderboard is now live: a dedicated page that tracks which models are writing code you’d actually merge. All scores — including Grok 4.5 and Inkling — are available, along with full methodology and sample tasks.
it's actually hilarious and sad that Anthropic who were once the goats of reasoning efficiency are now just slop-maxxing big fan of whatever OpenAI is doing
FrontierCode by Cognition now has a live leaderboard btw https://cognition.com/frontiercode
Grok has the best value for coding
@inductionheads don't ask me I think it's fine that top right is better
See the scores: https://cognition.com/frontiercode
Positive users praise Cognition's FrontierCode leaderboard for tracking merge-worthy AI code over synthetic benchmarks while negative users call the charts hard to read and insult models like Sonnet 5.
Based on 29 visible X reactions from 67 accounts; directional sample.
Ask a question below.
Published answers will appear here.