OpenAI Releases GPT-6 Sol and Luna Models

Model Performance and Cost Efficiency

OpenAI has introduced two new models, GPT-6 Sol and Luna, which prioritize cost-efficiency while maintaining high performance for agentic workflows. These models are positioned as workhorse alternatives to the flagship GPT-6 Astra. On the Automation Bench, GPT-6 Sol achieves a score of 33% at max effort, trailing the top-tier Astra model but significantly outperforming previous iterations at a lower price point. GPT-6 Luna, while scoring lower on general benchmarks, demonstrates high utility for its price, achieving a 66.6% score on the Frontier Code benchmark at a cost of 22 cents per task.

Pricing and Availability

OpenAI has implemented a 50% permanent price reduction for both models compared to their predecessors. GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is priced at 10 cents per million input tokens and 50 cents per million output tokens. Both models are currently available to ChatGPT Plus, Pro, Business, Enterprise, and Education users, with free users gaining access to GPT-6 Luna via the desktop application.

Strategic Market Positioning

The release of these models follows a trend of companies prioritizing efficiency over raw intelligence after initial frontier model launches. By providing distilled or architecturally optimized versions of their flagship models, companies like OpenAI and Anthropic are lowering the barrier for agentic coding and automation tasks. This shift suggests a broader industry move toward making high-performance models viable for production use cases where cost-per-task is a primary constraint.