OpenAI cuts GPT-6 prices in half with Sol and Luna
OpenAI (read this guide) has released GPT-6 Sol and GPT-6 Luna, cutting API prices by half compared to the GPT-5.6 versions.
Pricing
- Sol: $2 per million input tokens, $10 per million output tokens.
- Luna: $0.10 per token, $0.50 per task.
These are 50% lower than current promotional pricing for GPT-5.6. Both models are available in ChatGPT Work and Codex from Tuesday, covering Plus, Pro, Business, Enterprise, and Edu accounts. Free and Go users get Luna in the desktop app.
What they're for
- Sol: Built for complex tasks including coding.
- Luna: Designed for high-volume jobs with clear goals like summarizing documents or answering short questions.
Additional Features
OpenAI also improved caching, offering a 90% discount on cached input-token reads and enhanced control over tool switching within prompts.
Performance Comparisons
The announcement highlights comparisons primarily based on cost per task against Anthropic's Claude:
-
AutomationBench tests business workflows across 47 tools. GPT-6 Sol achieved a score of 33.2% at $0.27 per task, surpassing Claude Opus 5 (26.9%) at 11.1 times the cost.
-
On Agents' Last Exam, Sol scored 56.4%, beating Opus 5's best result at 60% lower cost per task.
-
DeepSWE software engineering test: Sol scored 68.8%, comparable to Opus 5 and Fable 5 at medium effort while costing 80% less.
-
OSWorld computer use: Sol achieved a score of 60.5% at extra-high effort, outperforming Opus 5 at medium effort (60.3%) with roughly 80% lower cost.
OpenAI notes some competitor scores were from published reports and certain comparisons involved indirect data. They also claim improved factual reliability based on internal tests using de-identified ChatGPT conversations.