ainewsblitz.com

Breaking

OpenAI's Codex Quietly Downgrades GPT-5.6 Sol Mid-Task, Developers Report

  • Foundation Models
  • AI Agents
  • Software Dev & Coding

Developers building on OpenAI's Codex say the platform has been silently dropping its flagship GPT-5.6 Sol model to lower effort tiers without notice, forcing them to manually restore the top "Ultra" setting to complete complex work. The behavior surfaced in a widely discussed example involving a 3D character animation pipeline, where the model's capability visibly collapsed each time it defaulted back to Light or Medium effort.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 5,714 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year