GPT-6.1 Sol lands on Merge Gateway

This makes GPT-6.1 Sol a compelling default for production coding agents, though developers should validate benchmark claims against their own repositories.

• –Near-Astra capability at lower reasoning and inference costs could materially improve agent economics
• –Merge Gateway adds unified access, routing, failover, observability, and spend controls across model providers
• –Merge’s Pro plan adds a 5% fee, so the one-fifth price claim is not fully all-in
• –DeepSWE results are promising but remain benchmark-specific; real-world codebase testing should determine model choice
• –Cached input costs just $0.10 per million tokens, making repeated agent context substantially cheaper