GPT-5.6 Trio and ChatGPT Work reshape how you tackle multi-step AI tasks

OpenAI rolled out three GPT-5.6 models—Sol (flagship, paid), Terra (balanced, accessible to free/Go users inside ChatGPT Work and Codex), and Luna (fastest)—along with ChatGPT Work, an agent for longer, multi-step tasks available in the unified desktop and mobile ChatGPT app. Users can pair models with different reasoning levels, which affects speed and cost, and pricing is token-based per model (e.g., Sol, Terra, Luna have distinct input/output costs), making the overall cost landscape more complex even as the workflow becomes more integrated. The goal is a set‑and‑forget experience, but users still need to pick the right model and settings for their tasks.
- ChatGPT's paradox of choice Axios
- GPT-5.6: Frontier intelligence that scales with your ambition OpenAI
- OpenAI, Meta, SpaceXAI Compete for More Cost-Efficient AI Models Bloomberg.com
- OpenAI Just Declared ChatGPT as You Know It Dead inc.com
- OpenAI's newest AI model is 54% more token efficient on agentic coding, Altman tells CNBC CNBC
Reading Insights
1
1
2 min
vs 3 min read
80%
510 → 101 words
Want the full story? Read the original article
Read on Axios