GPT-6 Sol and Luna halve OpenAI's prices
Artificial Analysis's evaluation finds GPT-6 Sol and Luna cost about half as much per task as GPT-5.6, with lower hallucination rates but no overall intelligence gain.
Artificial Analysis has published independent benchmarks of OpenAI's GPT-6 Sol and Luna. It finds that both models score the same as their GPT-5.6 predecessors on its Intelligence Index and Coding Agent Index, at roughly half the cost.
Pricing. Sol drops from $4/$20 to $2/$10 per million input/output tokens. Luna drops from $0.20/$1.20 to $0.10/$0.50. Cache terms are unchanged: reads get a 90% discount and writes carry a 25% premium.
Cost per task. Running the Intelligence Index costs $1.06 per task on GPT-6 Sol (max), down from $1.99. Luna (max) costs $0.07, down from $0.18. The savings come from the price cut alone, because both models use slightly more output tokens per task: 31k vs 29k for Sol and 51k vs 41k for Luna.
Coding. In OpenAI's Codex harness, Sol scores 57 on the Coding Agent Index, up 2 points. It gains on Terminal-Bench 4.0 (43% vs 37%) and SWE-Atlas-QnA (58% vs 54%), and its $2.99 cost per task puts it on the cost-performance frontier. Luna falls 2 points to 41.
Hallucination. On AA-Omniscience, Sol's hallucination rate falls from 92% to 60%, and Luna's falls from 93% to 77%. Sol gets there mainly by declining to answer more often. It attempts 83% of questions, down from 99%, and its accuracy drops from 59% to 54%.
Regressions. On GDPval-AA v2.1, a set of economically valuable tasks, Sol loses about 100 Elo points and Luna about 75. Luna also loses about 45 Elo points on AA-Briefcase v1.1. Artificial Analysis manually inspected the outputs and found the drops come from weaker presentation and deliverables that leave out required rubric elements.