Member-only story
GPT-6 Soul & Luna Just Dropped — Here’s How They Actually Fit Next to Opus 5.5
We got three new model recently.
About an hour after Anthropic released Claude Opus 5.5, OpenAI dropped GPT-6 Soul and GPT-6 Luna. Opus 5.5 looked strong on both benchmarks and cost. So the natural question is: what do these two new GPT-6 variants actually bring to the table?
Understanding the GPT-6 Family
The GPT-6 lineup currently has three tiers:
GPT-6 Astra
- Role: Flagship
- Positioning: Maximum capability
GPT-6 Soul
- Role: High-efficiency mid-tier
- Positioning: Near-Astra performance, much cheaper
GPT-6 Luna
- Role: Ultra-efficient lower tier
- Positioning: Strong enough for most everyday work
Soul and Luna are not trying to beat Astra. They are designed to deliver close performance at a significantly lower cost for tasks that don’t need the absolute top model.
Judging them against Astra and calling them “worse” misses the point. The right comparison is cost-adjusted performance on work that is only moderately complex.








