OpenAI expanded its GPT-6 lineup on Tuesday with Sol and Luna, two models priced at roughly half of their GPT-5.6 predecessors. GPT-6 Sol lands at $2 per million input tokens and $10 per million output tokens, down from $4 and $20. GPT-6 Luna drops to $0.10 input and $0.50 output, a 50% cut on input and about 58% on output. An OpenAI spokesperson told VentureBeat the new rates are permanent rather than promotional. Batch and Flex processing run at half those rates again, putting Sol output at $5 and Luna output at $0.25 per million tokens.
The benchmarks cut both ways. Artificial Analysis reports GPT-6 Sol (max) scoring 57 on its Coding Agent Index, up 2 points from GPT-5.6 Sol (max), with Terminal-Bench 4.0 improving from 37% to 43% and SWE-Atlas-QnA from 54% to 58%. At $2.99 per task it costs roughly half of GPT-5.6 Sol (max) and sits on the Pareto frontier of capability versus cost. GPT-6 Luna (max) is a different story: it scores 41, down 2 points, with SWE-Atlas-QnA falling to 44% and DeepSWE v1.1 to 64%, but arrives at about 60% lower cost per task. Luna ships with a newer knowledge cutoff than Sol.
Timing made the launch a head to head. Anthropic released Opus 5.5 about 90 minutes earlier, claiming 40% lower cost than Opus 5 at $4 per million input and $20 output. GPT-6 Sol matches Claude Sonnet 5 pricing at $2 and $10, undercutting Opus 5.5 by half, while Luna at $0.10 input lands in hosted open weight territory. Hacker News commenters focused on the price floor, with one noting Luna is now cheaper than DeepSeek and another asking how competing providers can serve comparable quality at that rate. OpenAI says paid ChatGPT plans get both models in Work and Codex today, and free users get Luna on desktop.