Summary
Highlights
Claude Opus 4.7, Antropic's latest AI model, shows significant improvements in programming. Tested against the SWBch Pro, a challenging AI benchmark, Opus 4.7 scored 64.3%, outperforming previous versions and competitors like Gemini 3.1 Pro.
Opus 4.7 demonstrates strong visual understanding, particularly in interpreting graphs and executing financial tasks, surpassing its predecessor, Opus 4.6. However, it struggles with memory retention, as evidenced by a low score in the MRCR version 2 'needle in a haystack' test.
While Opus 4.7 shows weaknesses in memory and internet browsing capabilities (Browsecom test), its overall AI intelligence, as measured by Arcagi 2, is comparable to Gemini 3.1 Pro. However, this intelligence comes at a significantly higher cost due to increased token usage leading to a 30% to 150% rise in task costs.
Anthropic has experienced unprecedented growth, with its annual run rate increasing from $15 billion to $30 billion in just two months. Despite this, the company faces criticism regarding its revenue reporting from Open AI's Chief Revenue Officer and struggles with frequent service outages due to scaling challenges and chip shortages.
The video highlights the broader issues facing the AI industry, such as the scarcity of specialized labor (e.g., plumbers, electricians for data centers) and the performance differences between leading AI models. Despite some frustrations with its speed, Claude Code remains a preferred tool for complex programming tasks due to its superior reasoning capabilities.