GPT-6.1 Terra (Max) placed eighth in Agent Arena's agent rankings

Agent Arena says GPT-6.1 Terra (Max) placed eighth in rankings drawing on nearly 9,000 agentic sessions. It placed third for Bash reliability and seventh for steerability, with a median task cost 43% below seventh-ranked Gemini 4 Argon (High). In a separate analysis, Arena says GPT-6.1 Sol (Max)'s cost advantage over Astra came mainly from cheaper tokens, while its advantage over Claude Opus came mainly from using fewer tokens per task.