GPT-6 Sol and Luna: OpenAI Cuts AI Costs in Half

Core News and Key Facts

The biggest change in OpenAI’s latest model launch is not simply the move from GPT-5.6 to GPT-6; it is how much AI capability the company is trying to make affordable. OpenAI has introduced GPT-6 Sol and GPT-6 Luna, expanding capabilities developed with GPT-6 Astra into lower-cost models aimed at professional work, coding, computer use and sustained AI assistance. OpenAI describes Astra as its highest-capability model, while Sol is designed to balance intelligence and cost and Luna focuses on efficient, high-volume workloads.

The API pricing is particularly notable. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna costs $0.10 and $0.50 respectively. OpenAI says these prices are 50% lower than the promotional pricing of their GPT-5.6 predecessors.

Context and Official Statements

OpenAI says the new models use training approaches similar to those behind Astra and improve areas including factuality, coding, computer use and alignment. Its published model information lists a 1.05-million-token context window and up to 128,000 output tokens for both Sol and Luna.

The company also reports improvements on several internal evaluations. GPT-6 Sol reached 33.2% on AutomationBench at xhigh effort, while GPT-6 Luna improved by 5.4 percentage points over GPT-5.6 Luna at high effort, according to OpenAI’s launch information. These are vendor-reported benchmark results and should be interpreted within the methodology of each test.

Availability is also being broadened. Sol and Luna are available through the API and in ChatGPT Work and Codex for eligible paid, business and education plans, while Free and Go users can access Luna through the desktop app. OpenAI says the models are being rolled out gradually in standard ChatGPT.

Krihaa Analysis

The strategic shift is clear: AI competition is moving from simply asking which model is smartest to asking how cheaply useful intelligence can be delivered at scale.

For developers, the 50% price reduction matters because the economics of an AI product can change dramatically when every automated task, coding cycle or agent interaction becomes cheaper. A lower per-token cost can make it practical to run AI more frequently instead of reserving advanced models for expensive, occasional tasks.

That could be particularly relevant for Indian startups and software teams, where controlling infrastructure and API costs is often as important as benchmark performance. Sol and Luna therefore represent more than two additional model names in OpenAI’s lineup. They show a push toward making stronger reasoning and agentic capabilities economically viable for everyday production workloads. The important test now is whether real-world reliability and developer adoption match the efficiency story OpenAI is presenting.