SWE-1.7 Closes Gap to High-End LLMs on Software Engineering Tasks

Cognition's SWE-1.7 now performs near GPT-4.5 and Claude Opus levels on software engineering benchmarks, suggesting specialized models can match frontier LLMs on narrow domains. For indie devs, this signals that task-specific AI tooling may offer better value than relying on general-purpose models for code tasks.