Skip to content

Back to the catalog

openai.comarticle

Introducing GPT-6 Sol and Luna

OpenAI introduces GPT-6 Sol and GPT-6 Luna, cheaper models trained with methods similar to GPT-6 Astra. API prices drop 50% against GPT-5.6 promotional pricing. The post reports cost-per-task comparisons against Claude Opus 5 and Fable 5/5.1: AutomationBench, Agents' Last Exam, FrontierCode, DeepSWE v1.1 (Sol 68.8%, Luna 66.6%) and OSWorld 2.0 offline. It also covers improved GPT-6 prompt caching (90% discount on cached reads, a dashboard, a diagnostics tool, explicit breakpoints, changing reasoning effort or tools without breaking the cache), factuality and alignment gains, and availability in ChatGPT Work, Codex and the API as `gpt-6-sol` and `gpt-6-luna`.

TL;DR OpenAI releases GPT-6 Sol and Luna, cost-efficient siblings of Astra, with 50% lower API prices, vendor-reported benchmark wins on cost per task, and better prompt caching.
Open original ↗

Organisations named here

  • OpenAI
  • Anthropic
  • GitHub
  • Agents' Last Exam
  • AutomationBench
  • DeepSWE
  • FrontierCode
  • OSWorld 2.0

Technologies named here

  • Codex
  • GPT-6 Sol
  • GPT-6 Luna
  • ChatGPT Work
  • GPT-6 Astra
  • OpenAI API
  • GPT-6.1 Sol
  • Prompt Caching Dashboard
  • Prompt caching diagnostics tool
  • GPT-5.6 Sol
  • Claude Opus 5
  • GitHub Copilot
  • Claude Fable 5.1
  • Claude Fable 5
  • ChatGPT
  • Opus 5

Techniques named here

  • Prompt caching
  • Explicit cache breakpoints
  • Changing reasoning effort and tools mid-conversation

Related in models