No AI agent scores above 12 percent on SemiAnalysis' new real-world work benchmark
SemiAnalysis is putting AI agents through simulated knowledge work in its new GDPx benchmark, covering everything from chip design to quant trading. No model breaks 12 percent. Claude Opus 5.5 leads narrowly over Gemini 4 Carbon. Critics point to the small number of tasks per…
AI Summary & Key Takeaways
- SemiAnalysis is putting AI agents through simulated knowledge work in its new GDPx benchmark, covering everything from chip design to quant trading
- No model breaks 12 percent
- Claude Opus 5.5 leads narrowly over Gemini 4 Carbon
No AI agent scores above 12 percent on SemiAnalysis' new real-world work benchmark
Related stories
Claude can now analyze your spending with its new Personal Finance tool
With Personal Finance in Claude, users can connect their bank and card accounts through Plaid and ask questions about their finances.
The world's best model for cybersecurity tasks is here!
The world's best model for cybersecurity tasks is here! Ladies and gentlemen, welcome GPT-6 Cyber. Judging by the published results, the model sets a new record in cybersecurity tasks. She scored 48.5% in the ExploitGym benchmark, discovered 81 previously unknown...
How Mozilla strengthens Firefox security with GPT-6 Cyber
Mozilla used GPT-6 Cyber to uncover previously unknown high-severity vulnerabilities in Firefox and identify architectural improvements. Its team shipped fixes to millions of users, turning findings into protections for the people who depend on the software. Hear Brian…
To catch up, OpenAI released GPT-6 Cyber. The model is capable of independently finding and exploiting vulnerabilities
To catch up, OpenAI released GPT-6 Cyber. The model is capable of independently finding and exploiting vulnerabilities. OpenAI has presented a specialized model for cybersecurity. GPT-6 Cyber can search for unknown vulnerabilities, analyze code and develop exploits for...