OpenAI Launches GPT-6 Sol and Luna With 50% Lower API Prices60 characters

OpenAI has launched GPT-6 Sol and GPT-6 Luna, two lower-cost models in its GPT-6 family. The company says API prices are 50% lower than the promotional pricing for their GPT-5.6 predecessors, while the new models bring improvements in coding, computer use, factuality, and agent workflows.

GPT-6 Sol and Luna Cut API Prices

The biggest change is pricing. OpenAI lists GPT-6 Sol at $2 per 1 million input tokens and $10 per 1 million output tokens. GPT-6 Luna costs $0.10 per 1 million input tokens and $0.50 per 1 million output tokens. OpenAI says both reductions are measured against GPT-5.6 promotional pricing.

• GPT-6 Sol: $2 input and $10 output per 1 million tokens
• GPT-6 Luna: $0.10 input and $0.50 output per 1 million tokens
• Cached input reads receive a 90% discount
• API model IDs are gpt-6-sol and gpt-6-luna

That puts the new models directly into a part of the market where API cost can determine whether developers run an AI agent continuously, use more reasoning, or process larger volumes of work. Saganote's AI API pricing guide provides broader context on how model costs compare across providers.

What GPT-6 Sol and Luna Add

OpenAI says Sol and Luna were trained with methods similar to GPT-6 Astra and carry over improvements in professional work, factuality, coding, computer use, and alignment. Sol is positioned for harder work where developers want more usage at a lower cost, while Luna is designed for focused, high-volume workloads.

OpenAI reports that GPT-6 Sol scored 68.8% on DeepSWE v1.1 at maximum effort, within 1.1 percentage points of Claude Fable 5's highest reported score in that evaluation. OpenAI says Sol reached that result at approximately 80% lower cost per task. These are OpenAI's reported evaluations, and the company notes that competitor results come from publicly available reports.

For computer-use tasks, OpenAI says GPT-6 Sol reached 60.5% on OSWorld 2.0 offline at xhigh effort, compared with 60.3% for Claude Opus 5 at medium effort, at approximately 80% lower cost per task. GPT-6 Luna also improves over its predecessor in OpenAI's testing.

The release follows OpenAI's earlier GPT-6 Astra launch, which introduced the company's higher-end GPT-6 model. It also builds directly on the pricing and model structure readers saw with GPT-5.6 Sol and Luna.

Prompt Caching Gets a Major Update

OpenAI is also changing how GPT-6 handles repeated context. The company says improved prompt caching delivers higher cache hit rates by default, while cached input-token reads receive a 90% discount. Developers can adjust reasoning effort and tool availability without breaking earlier cache context, and explicit cache breakpoints provide more control over reusable prefixes.

OpenAI says GitHub has seen the share of prompt tokens requiring fresh processing fall by more than 50% across billions of requests. That matters most for long-running agents and coding workflows where the same context can be reused across many calls.

GPT-6 Arrives as AI Model Prices Keep Falling

The launch also lands on the same day Anthropic introduced Claude Opus 5.5. Anthropic says that model costs 40% less to run than Opus 5 on typical workloads, with $4 input and $20 output pricing per 1 million tokens. The same-day releases put renewed attention on the cost of running advanced coding and agent workloads.

For readers tracking the broader pricing shift, Saganote's ChatGPT vs. Claude pricing comparison and OpenAI vs. Anthropic vs. Google DeepMind overview provide additional context. OpenAI's earlier Grok 4.6 pricing launch is another relevant example of lower-cost model competition.

Availability

GPT-6 Sol and GPT-6 Luna are available in the OpenAI API as gpt-6-sol and gpt-6-luna. In ChatGPT, both models are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. OpenAI says the models are not yet available in Chat and are rolling out gradually in ChatGPT Work and Codex.

The company is positioning GPT-6 Astra as its highest-end model, while Sol and Luna extend GPT-6 capabilities to lower-cost workloads. Developers can now choose between the three tiers based on the amount of capability and cost their application requires.