OpenAI expands GPT-6 lineup with cheaper, faster Sol and Luna models
OpenAI has released two new models for different needs. ChatGPT-maker OpenAI has now introduced GPT-6 Sol and GPT-6 Luna models. The company claims that Sol is designed as a high-performance model optimised for complex reasoning, large-scale research and enterprise applications. On the other hand, Luna, is a lighter and more adaptive model which is built for everyday use cases and offer faster responses and lower compute requirements. Both the new models are created with OpenAI’s latest alignment techniques, including safeguards around recursive self-improvement (RSI) and improved oversight mechanisms. OpenAI also stressed on the fact that the launch reflects its commitment to balancing innovation with safety, ensuring that frontier AI systems remain under human control while scaling their capabilities.
Cheaper models built on the same foundationEarlier this month, OpenAI introduced GPT-6 Astra, which it describes as the most intelligent and aligned model it has built to date. Sol and Luna extend that same generation of intelligence downward, trained using similar methods as Astra but optimized to advance the frontier on cost efficiency rather than raw capability. OpenAI says the two new models bring Astra's advances in professional work, factuality, coding, computer use, and alignment to faster, more affordable tiers.
The economics are the headline change here. Thanks to improvements in caching and inference, OpenAI is cutting API prices for both models by 50% compared to their GPT-5.6 promotional pricing. GPT-6 Sol now costs $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively. GPT-6 Luna is even cheaper, dropping to $0.10 per million input tokens and $0.50 per million output tokens, down from $0.20 and $1.20. Astra remains OpenAI's top recommendation when uncompromising performance matters most.
How the new models performOpenAI is positioning Sol in particular as a serious contender against rival models at a fraction of the cost. On AutomationBench, a test of real-world business workflows, the company says GPT-6 Sol at its highest effort setting outperforms Anthropic's Claude Opus 5 at max effort while costing just 9% as much per task. On Agents' Last Exam, which evaluates complex, long-horizon professional workflows, Sol scored 56.4% at max effort — ahead of Opus 5's best score in the same evaluation, at 60% lower cost per task.
Coding performance also saw a notable jump. On FrontierCode, which grades AI-generated code on whether it's actually ready to merge into real codebases, OpenAI says Sol improved substantially over its GPT-5.6 predecessor and now matches Anthropic's Claude Fable 5.1 at a much lower price. On a separate software-engineering benchmark, DeepSWE, Sol scored within roughly one percentage point of Fable 5's best result while costing about 80% less per task.
Factuality — long a weak point for large language models — also improved meaningfully. Based on OpenAI's internal evaluation using real conversations where users had previously flagged mistakes, Sol now makes roughly half as many factual errors as its predecessor, approaching Astra-level reliability at a fraction of the cost.
Key features* Sol: Advanced reasoning, multi‑step task automation, and integration with scientific and enterprise workflows.
* Luna: Lightweight deployment, optimized for consumer applications, and faster performance on smaller devices.
* Shared Advances: Improved multilingual support, adaptive memory, and enhanced contextual understanding across domains.
AvailabilityGPT-6 Sol and Luna are rolling out today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, while Free and Go users can access Luna through the desktop app. Neither model is yet available in standard Chat. Developers can access both through the OpenAI API as gpt-6-sol and gpt-6-luna. OpenAI said the rollout within ChatGPT will happen gradually throughout the day to keep the service stable.
Cheaper models built on the same foundationEarlier this month, OpenAI introduced GPT-6 Astra, which it describes as the most intelligent and aligned model it has built to date. Sol and Luna extend that same generation of intelligence downward, trained using similar methods as Astra but optimized to advance the frontier on cost efficiency rather than raw capability. OpenAI says the two new models bring Astra's advances in professional work, factuality, coding, computer use, and alignment to faster, more affordable tiers.
The economics are the headline change here. Thanks to improvements in caching and inference, OpenAI is cutting API prices for both models by 50% compared to their GPT-5.6 promotional pricing. GPT-6 Sol now costs $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively. GPT-6 Luna is even cheaper, dropping to $0.10 per million input tokens and $0.50 per million output tokens, down from $0.20 and $1.20. Astra remains OpenAI's top recommendation when uncompromising performance matters most.
How the new models performOpenAI is positioning Sol in particular as a serious contender against rival models at a fraction of the cost. On AutomationBench, a test of real-world business workflows, the company says GPT-6 Sol at its highest effort setting outperforms Anthropic's Claude Opus 5 at max effort while costing just 9% as much per task. On Agents' Last Exam, which evaluates complex, long-horizon professional workflows, Sol scored 56.4% at max effort — ahead of Opus 5's best score in the same evaluation, at 60% lower cost per task.
Coding performance also saw a notable jump. On FrontierCode, which grades AI-generated code on whether it's actually ready to merge into real codebases, OpenAI says Sol improved substantially over its GPT-5.6 predecessor and now matches Anthropic's Claude Fable 5.1 at a much lower price. On a separate software-engineering benchmark, DeepSWE, Sol scored within roughly one percentage point of Fable 5's best result while costing about 80% less per task.
Factuality — long a weak point for large language models — also improved meaningfully. Based on OpenAI's internal evaluation using real conversations where users had previously flagged mistakes, Sol now makes roughly half as many factual errors as its predecessor, approaching Astra-level reliability at a fraction of the cost.
Key features* Sol: Advanced reasoning, multi‑step task automation, and integration with scientific and enterprise workflows.
* Luna: Lightweight deployment, optimized for consumer applications, and faster performance on smaller devices.
* Shared Advances: Improved multilingual support, adaptive memory, and enhanced contextual understanding across domains.
AvailabilityGPT-6 Sol and Luna are rolling out today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, while Free and Go users can access Luna through the desktop app. Neither model is yet available in standard Chat. Developers can access both through the OpenAI API as gpt-6-sol and gpt-6-luna. OpenAI said the rollout within ChatGPT will happen gradually throughout the day to keep the service stable.
Next Story