This morning, I conducted an in-depth comparison test of GPT 5.6 Sol Ultra, Claude Fable 5, and Grok 4.5. I had the three models simultaneously perform a comprehensive audit of my existing system backend code (approximately 185k lines), then aggregated and scored the audit results. The conclusion is:



GPT 5.6 Sol Ultra >= Claude Fable 5, and much cheaper. The only downside is it's very slow — must enable 1.5x speed.

If Fable 5 is no longer open to regular users after 7/12, customers will vote with their feet.

Grok 4.5's model capability has basically caught up with the previous generation GPT 5.5 and Opus 4.8, showing rapid progress. Its biggest advantages are very low price and extremely fast speed — a $30 subscription is more than enough. Currently, xAI has the most computing power; Musk is still rich and generous.

I'll try Meta's new model Spark later.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned