OpenAI Launches GPT-5.6: New Features for ChatGPT and API

OpenAI launches GPT-5.6: Sun, Earth, and Moon for ChatGPT and API

OpenAI has finally released GPT-5.6 for general access. And it’s not just another update to speed up emails. We are talking about a radical redesign that transforms artificial intelligence from a mere chat tool into a true worker. In less than 24 hours, the model will be available to everyone. If you have an account on ChatGPT, ChatGPT Work, or Codex, get ready to notice the changes. The limited preview phase that frustrated us so much is behind us. Now, the model is available globally, promising greater intelligence per token and much lower latency.

Three flavors to dominate the market

Sam Altman and his team decided to segment the technology into three clear levels: Sun, Earth, and Moon. Sun is the main model, designed for the most demanding coding and scientific tasks. Earth is the balanced option for everyday use. And at the base, we have Moon, the ultra-fast alternative for when budget and time are tight. The prices reflect this segmentation. Sun costs $5 per million input tokens and $30 for output. Earth cuts that fee in half, while Moon reduces costs even further. A strategy that makes perfect sense for startups.

Prompt cache and "Ultra Mode"

Additionally, OpenAI has adjusted the prompt cache system. Now, it is much more predictable, with explicit breakpoints and a minimum duration of 30 minutes in memory. Reading from the cache still saves 90% of the cost compared to the base rate, although writing costs a bit more. But the real cherry on top of GPT-5.6 is the ultra adjustment. This new mode activates up to four agents working in parallel on the same request, dividing the work to solve complex problems. And it doesn’t stop there: the model can now write and execute JavaScript directly in its own memory, coordinating external tools without needing to return the code first.