“Meter's primary focus for benchmarking is on tasks an engineer at a frontier AI lab might perform, such as software engineering and fine-tuning AI models.”