OpenAI Launches GPT-6 Sol and Luna With 50% Lower API Prices

In addition to the launch of Anthropic’s Claude Opus 5.5 and SpaceXAI’s Grok 4.7, OpenAI has expanded its latest AI model lineup with GPT-6 Sol and GPT-6 Luna. Both the new models that are stronger at coding, professional work and everyday AI tasks. The launch follows GPT-6 Astra, which released earlier this month as OpenAI’s flagship model.

GPT-6 Sol and Luna focus on capability and lower costs

OpenAI says both models were developed using methods similar to those behind GPT-6 Astra, with improvements in professional work, coding, computer use, factuality and alignment.

GPT-6 Sol is positioned as the more capable option of the two, while Luna focuses on delivering faster and cheaper access to the same generation of technology. OpenAI says improvements to caching and inference allow it to reduce API prices by 50% compared with the promotional pricing of GPT-5.6 Sol and Luna.

The new pricing puts GPT-6 Sol at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens.

OpenAI is also pushing coding and agentic workloads

Coding is another major focus for the new models. OpenAI says GPT-6 Sol performs substantially better than GPT-5.6 Sol on its FrontierCode evaluation, while GPT-6 Luna also delivers stronger coding performance at a lower cost.

The company says GPT-6 Sol reaches 68.8% on DeepSWE v1.1 at maximum effort, while Luna scores 66.6%. OpenAI also highlights improvements in computer-use tasks, with Sol achieving a similar OSWorld 2.0 score to Claude Opus 5 at medium effort in its comparison.

Both models are also being positioned for AI agents and longer-running workflows. OpenAI says the new models inherit improvements from Astra that should make conversations clearer, reduce unnecessary jargon and produce fewer odd turns of phrase.