qwen 3.8 max vs deepseek v4 flash 0731 vs kimi k3 vs gpt 5.6 sol – on rubik's cube and chess
seedance 2.5 vs minimax h3 vs grok imagine 1.5 vs gemini omni
thinking machines' inkling-small fixed more bugs than inkling – at 3.5x fewer parameters and 11x cheaper
92% of ai hacking agents fall for a cybersecurity trap – 2.5x more often than a human hacker would
top five ai rounds from july – tracked by our ai host nathan on capital radar:
qwen 3.8 max outperforms every us frontier model on agentic tasks by 4x+ on price/quality
buzz alone pulled 9k stars in seven days, +76% growth
buzz just pulled 10.5k github stars in a week – +111% growth – so we set one up from scratch and did a full review
this month openrouter is chinese. two of the top 10 most-used models cost $0
openai cut gpt-5.6 luna api pricing by 80%. here's where that leaves the 50-51 band on artificial analysis