OpenAI has expanded its GPT-6 lineup with two new models, GPT-6 Sol and GPT-6 Luna, cutting API prices in half compared to their GPT-5.6 predecessors. Both models bring the training methods behind GPT-6 Astra down to cheaper, faster tiers.
The announcement comes three weeks after the company launched GPT-6 Astra, which remains OpenAI’s top model overall. GPT-6 Sol and GPT-6 Luna are built to bring most of that intelligence to everyday, budget-conscious work instead.
GPT-6 Sol and Luna Cost Half of What Their Predecessors Did
OpenAI cut API pricing for both models by 50 percent compared to GPT-5.6 promotional rates:
- GPT-6 Sol: $2 per million input tokens (down from $4), $10 per million output tokens (down from $20)
- GPT-6 Luna: $0.10 per million input tokens (down from $0.20), $0.50 per million output tokens (down from $1.20)
OpenAI also improved prompt caching for both models, pushing cached input-token reads to a 90 percent discount and raising cache hit rates by default. GitHub told OpenAI the changes cut the share of prompt tokens needing fresh processing by more than 50 percent across billions of requests, which is part of what speeds up Copilot.
GPT-6 Sol Beats Claude Opus 5 at a Fraction of the Cost
On AutomationBench, a test of business workflows across connected apps, GPT-6 Sol at its highest effort setting outscores Claude Opus 5 at its own highest effort setting while costing about 9 percent as much per task. On Agents’ Last Exam, which evaluates long, complex professional workflows, GPT-6 Sol scores 56.4 percent, above Claude Opus 5’s best result at 60 percent lower cost per task.
Coding numbers follow the same pattern, on DeepSWE v1.1, GPT-6 Sol scores 68.8 percent, within 1.1 points of Claude Fable 5’s best result, at roughly 80 percent lower cost per task. GPT-6 Luna scores 66.6 percent on the same benchmark, comparable to Claude Opus 5 and Claude Fable 5 at medium effort, while costing 93 percent less than Claude Opus 5 and 96 percent less than Claude Fable 5 per task.
On factuality, measured against real ChatGPT conversations where users had flagged a mistake, GPT-6 Sol makes about half as many errors as GPT-5.6 Sol. GPT-6 Luna improves enough at higher effort settings to match GPT-5.6 Sol’s accuracy at about a hundredth of the cost.
Both Models Ship Today, but Not Yet in ChatGPT’s Free Chat Tier
GPT-6 Sol and GPT-6 Luna are rolling out today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu accounts. Free and Go users can reach GPT-6 Luna through the desktop app. Neither model is available in regular ChatGPT chat yet. In the API, they’re listed as gpt-6-sol and gpt-6-luna.
