OpenAI Releases GPT-6 Sol and Luna: 50% Cheaper API Pricing and Benchmarks
What changed
OpenAI launched two new models, GPT-6 Sol and GPT-6 Luna, built on similar training methods as their recent GPT-6 Astra. These two models are priced significantly lower than earlier GPT-6 versions, with Sol costing $2 per 1 million tokens for prompt usage and $10 for completion, and Luna at an even cheaper $0.10 and $0.50 respectively. Both are available immediately via OpenAI’s API, as well as integrated into ChatGPT Work and Codex environments.
Why builders should care
The cost drop by around 50 percent removes a key barrier to scaling AI-powered applications that rely on frequent or high-volume interactions. Less expensive models mean startups, agencies, and developers can run larger workloads or experiment more aggressively without blowing budgets. The availability of these models in ChatGPT Work and Codex also encourages tighter integration of AI agents into workflows involving natural language and code generation.
The practical takeaway
Builders looking to deploy customer bots, automated agents, or development helpers can now choose lower-price GPT-6 variants without completely sacrificing the underlying training quality of GPT-6 Astra. The improved prompt caching also optimizes performance for long-running agents, reducing latency and compute overhead in sustained sessions. This can directly impact usability and economics in real-world AI deployments.
What to watch next
It will be important to track how OpenAI prices and positions these models relative to flagship GPT-6 versions going forward. The company may further accelerate cost reductions or push these lower-cost models into more niche, cost-sensitive use cases. Monitoring performance benchmarks will reveal if cheaper pricing brings real-world tradeoffs in output quality. Also, watch for competitor responses and how customers adjust their AI strategy incorporating these cheaper, purpose-built GPT-6 options.
AI Quick Briefs Editorial Desk