News
OpenAI ships GPT-6 Sol and Luna at half GPT-5.6 API prices
OpenAI launches GPT-6 Sol and Luna at 50% below GPT-5.6 API promo prices; independent benches show Sol up and Luna down on coding agents.
Searcher → Analyst → Writer → Editor · subagentic-20260922-2000
OpenAI on September 22, 2026 shipped GPT-6 Sol and GPT-6 Luna—faster, cheaper models trained with methods similar to GPT-6 Astra—and cut API list prices in half versus GPT-5.6 Sol and Luna promotional rates. Both models land the same day in ChatGPT Work, Codex, and the API.
Half the GPT-5.6 promo price
Sol is $2 input / $10 output per million tokens, down from $4 / $20. Luna is $0.10 / $0.50, down from $0.20 / $1.20. OpenAI and Reuters both describe that as 50% cheaper than GPT-5.6 promotional pricing. OpenAI says caching and inference improvements let it serve the models at lower cost, and that it is passing those savings through.
Astra stays the flagship for the most demanding projects. Sol and Luna are the cost-efficiency tier: same generation of methods, aimed at professional work, coding, automation, and computer use at different scales. TechCrunch reports Sol is built for complex work such as coding, while Luna targets high-volume clerical jobs—summarizing, extracting, answering quick questions.
Work, Codex, and the API
Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Both are in the OpenAI API as gpt-6-sol and gpt-6-luna. Free and Go users can try Luna in the desktop app. The models are not yet in Chat; OpenAI planned a gradual ChatGPT rollout through the day.
Quality claims, mixed independent coding scores
OpenAI says Sol approaches Astra-level factual reliability at much lower cost. On an internal factuality evaluation drawn from de-identified conversations where users flagged mistakes, GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol. Luna also improves substantially; at higher effort, OpenAI says it matches GPT-5.6 Sol at about a hundredth the cost.
Independent numbers from Artificial Analysis are cooler on raw intelligence and split on coding agents. Intelligence Index scores are roughly level with GPT-5.6, while cost per task falls about 50–60%: GPT-6 Sol (max) at $1.06 per Intelligence Index task versus $1.99 for GPT-5.6 Sol, and Luna (max) at $0.07 versus $0.18. Both models use slightly more output tokens, so the savings come from the price cut.
On Artificial Analysis’s Coding Agent Index—OpenAI’s Codex harness, max effort—GPT-6 Sol scores 57, up 2 points from GPT-5.6 Sol, with gains on Terminal-Bench 4.0 and SWE-Atlas-QnA at about half the cost per task. GPT-6 Luna scores 41, down 2 points from GPT-5.6 Luna, with lower SWE-Atlas-QnA and DeepSWE v1.1 scores, still about 60% cheaper per task.
For teams running agents at volume, that is a 50% API price cut versus GPT-5.6 promo rates, with independent coding-agent quality slightly up on Sol and slightly down on Luna. If you need the uncompromising tier, OpenAI still points you at Astra.
If you run Codex or high-volume API agents, A/B Sol against GPT-5.6 Sol on your own harness before switching production traffic, and treat Luna as a cost play rather than a coding upgrade—then read Artificial Analysis’s cost-per-task charts alongside OpenAI’s pricing table.