OpenAI Introduces ChatGPT 5.6 Soul, Terra and Luna Models
TL;DR
OpenAI has reportedly introduced GPT-5.6 with three variants: Soul for complex reasoning, Terra as a general-purpose option and Luna for fast, large-scale processing. The report points to Soul Ultra, sub-agent orchestration, Terminal Bench and Exploit Bench results, plus up to 750 tokens per second via Cerebras chips. It also flags risks around cybersecurity, bio-research misuse, hallucinations and restricted access through a controlled preview rather than broad availability.
Nauti's Take
If this is real, OpenAI isn't just shipping a faster model, it's building a fleet for different jobs. Builders should stop thinking one-size-fits-all and start thinking orchestration, benchmarks, and access control.
The bottleneck moves from prompting to system design.
Briefingshow
If the claims hold up, GPT-5.6 moves the story from one flagship model to a tiered family for reasoning, everyday work and throughput. The bigger issue is not the benchmark bragging, but access control: as models get stronger in cybersecurity and bio-research, deployment becomes a political product as much as a technical one.
Video
Sources