GPT-6 Sol and GPT-6 Luna have been introduced by OpenAI, and some capabilities of the GPT-6 Astra model will be available at lower tiers. These models are intended for professional tasks, programming, computing, and any other tasks where continuous assistance is required from the AI.
Furthermore, the pricing of the API service is halved compared to the promotional price of the previous versions of GPT-5.6 Sol and GPT-5.6 Luna.
According to OpenAI, the training procedure of GPT-6 Sol and GPT-6 Luna is identical to that of GPT-6 Astra. These new models provide better results when completing professional tasks, being factually correct, coding, computing, and aligning. The GPT-6 Astra model is still OpenAI’s best model.
GPT-6 Sol and GPT-6 Luna Availability
GPT-6 Sol and GPT-6 Luna are available in ChatGPT Work and Codex on the Plus, Pro, Business, Enterprise, and Edu subscriptions.
GPT-6 Luna is accessible to Free and Go users via the desktop application. The two models are not yet available in the default version of ChatGPT, although OpenAI is planning to introduce them in the coming time.
Developers have access to the two models through the API as gpt-6-sol and gpt-6-luna.
GPT-6 Sol and GPT-6 Luna API Pricing
GPT-6 Sol costs $2 (approximately Rs. 191) for one million input tokens and $10 (approximately Rs. 957) for one million output tokens via the API.
GPT-6 Luna comes at a cost of $0.10 (approximately Rs. 10) for one million input tokens and $0.50 (approximately Rs. 48) for one million output tokens.
According to OpenAI, both of these models are 50% cheaper than the promotional rates of GPT-5.6 Sol and GPT-5.6 Luna.
| Model | Price per 1 Million Input Tokens | Price per 1 Million Output Tokens |
|---|---|---|
| GPT-6 Sol | $2 (roughly Rs. 191) | $10 (roughly Rs. 957) |
| GPT-6 Luna | $0.10 (roughly Rs. 10) | $0.50 (roughly Rs. 48) |
GPT-6 Sol and GPT-6 Luna Performance
The two models are said to have enhanced performance in professional tasks, programming, computing, and accuracy.
On AutomationBench, GPT-6 Sol scored 33.2% at the high effort level. On the other hand, Claude Opus 5 had 26.9% at the highest effort level. The cost for GPT-6 Sol is reported to have been $0.27 (Rs. 26) per task on the test.
The performance on AutomationBench by GPT-6 Luna increased by 5.4% from GPT-5.6 Luna at the high effort level. In addition, the cost of the task was reduced by 58% by GPT-6 Luna.
On the DeepSWE benchmark, GPT-6 Sol had 68.8% at the maximum effort level. Claude Fable 5 also had 69.9%, but OpenAI claims that the cost per task was about 80% less.
GPT-6 Sol scored 60.5% on the OSWorld 2.0 benchmark at the high effort level. At the medium effort level, Claude Opus 5 scored 60.3%. The cost per task is reported to be 80% lower for GPT-6 Sol.
Improved Factuality and Clearer Responses
According to OpenAI, GPT-6 Sol commits only half the errors that GPT-5.6 Sol commits within its internal fact-checking test.
The firm also asserts that GPT-6 Luna performs at par with GPT-5.6 Sol in terms of effort but costs about one-hundredth as much.
According to OpenAI, GPT-6 Sol and GPT-6 Luna offer clear responses, and they employ minimal jargon and fewer details.
Explore the latest developments in artificial intelligence, advanced AI models, and emerging technology with SmartMag.
