1GPT-5.6 Sol
No. 1 — GPT-5.6 Sol. On the lists of 8 of 8 AIs; median place there: 3. It writes the cleanest code, with the fewest counted bugs, at a low output price, though it is weaker on day-long tasks.
Did you know
GPT-5.6 comes in three sizes: Sol, Terra and Luna. OpenAI's figures say the cheaper Terra and Luna beat Fable 5 on some tests at around one-sixteenth of the cost.
- What it is
- OpenAI's Sol coding model, sold in three sizes.
- Its edge
- Fewest counted bugs of the three, at a much lower output price.
- Who it suits
- Suits everyday clean coding; not the pick for 20-hour marathon tasks.
- Relative bug rate0.76 relative bug rate
History
On June 26, 2026, OpenAI previewed GPT-5.6 Sol only to a small group of vetted partners. At the US government's request, it shared the partner names with the government first. OpenAI called this a short-term step, not a permanent arrangement.
Where to go further
Open the GPT-5.6 Sol model page. It lists the model ID gpt-5.6-sol, a 1,050,000-token context window, and the price. Start with one demanding real task at the default medium effort, then raise or lower it.