Updated
Updated · TechCrunch · Sep 22
OpenAI Launches GPT-6 Sol and Luna at Half the Cost, Claiming Fewer Errors
Updated
Updated · TechCrunch · Sep 22

OpenAI Launches GPT-6 Sol and Luna at Half the Cost, Claiming Fewer Errors

3 articles · Updated · TechCrunch · Sep 22

Summary

  • GPT-6 Sol and Luna are now rolling out across ChatGPT Work, Codex and the API, with Luna also headed to the desktop app plus Free and Go tiers.
  • 50% lower pricing versus the 5.6-series models is the headline change, which OpenAI attributes to caching and inference improvements aimed at making GPT-6 cheaper and more accessible.
  • OpenAI said GPT-6 Sol makes about half as many factual mistakes as its predecessor on an internal evaluation and reaches Astra-level reliability at a lower cost, while also cutting coding errors.
  • The release expands the GPT-6 lineup after Astra launched earlier this month, and lands amid a tight race with Anthropic, which updated Opus 5.5 about 90 minutes before OpenAI's announcement.

Insights

As OpenAI slashes GPT-6 prices, is the AI industry entering a ruthless race to the bottom for enterprise dominance?
Will OpenAI's specialized Sol model truly beat Anthropic's Opus 5.5 in real-world coding, or only in internal tests?
Can smart model routing and caching actually halve enterprise AI costs, or will complex reasoning tasks secretly erase those savings?