APEX-Agents-AA: First Professional Services Benchmark Reveals AI Agents Fail 67% of Long-Horizon Tasks — GPT-5.4 Leads at 33.3%

Artificial Analysis·high signal

Artificial Analysis launched APEX-Agents-AA, evaluating AI agents on 452 real professional tasks spanning investment banking, management consulting, and corporate law with 5-10 day simulated engagements. Top scores: GPT-5.4 at 33.3%, Claude Opus 4.6 at 33.0%, Gemini 3.1 Pro at 32%. The ~33% ceiling reveals that traditional benchmark scores correlate poorly with real-world agent task completion — models scoring 90%+ on SWE-bench fail two-thirds of professional workflow tasks.