Claude Opus 5.5 vs GPT-6 Astra vs GPT-6 Sol: effort setting matters more than model

Artificial Analysis ran Claude Opus 5.5, GPT-6 Astra and GPT-6 Sol through the same ten evaluations at every reasoning-effort level. Opus 5.5 (max) scores 58 on the Intelligence Index vs 53 for Astra and 48 for Sol, but emits ~4x more output tokens, making it costlier per task at $5.98 vs $3.26. Below ~$1.30 per task Sol and Astra lead; above that Opus 5.5 wins.

• Opus 5.5 (max) scores 58 on AA Intelligence Index, Astra (max) 53, Sol (max) 48
• Opus 5.5 emits ~119k output tokens per task, Astra ~27k, Sol ~31k
• Under $1.30 per task Sol and Astra lead; above that Opus 5.5 wins
• In the same harness Opus 5.5 and Astra both scored 59.6% on Terminal-Bench 4.0