Key takeaways

• GPT-6 Luna is a cost-efficient and fast text+image+file-to-text model from OpenAI, part of the GPT-6 series, suited for high-volume, latency-sensitive tasks like chat, classification, summarization, information extraction, and quick Q&A.
• Released on September 22, 2026, it extends the GPT-6 Astra generation's intelligence at more accessible price points, featuring an MoE architecture with a context length of 1,050,000 and max tokens of 100,000.
• Significant cost improvements are noted, with the model being half the cost of GPT-5.6 Luna and offering a 90% discount on cached reads due to advancements in caching and inference.
• The model is proprietary, with no open weights, and is provided by OpenAI, also available through OpenRouter. It boasts very low costs per token: $1e-07 per input token and $5e-07 per output token.
• Its primary purpose is to deliver advanced GPT-6 capabilities for enterprise-level applications requiring efficiency and low latency, making cutting-edge AI more affordable for high-demand scenarios.

Technical specifications

Context length

1M tokens

Max output

—

Architecture

MoE

Modalities

text+image+file->text

Parameters

—

Open weights

No

Licence

Proprietary

Regions

USA

Vendor

OpenAI

Provider model key

openai/gpt-6-luna